the modern firm - department home · holmström, paul milgrom, and joel podolny. they will...

The Modern Firm

This page intentionally left blank

The ModernFirm

Organizational Design forPerformance and Growth

John Roberts

1

3Great Clarendon Street, Oxford OX2 6DP

Oxford University Press is a department of the University of Oxford.It furthers the University’s objective of excellence in research, scholarship,

and education by publishing worldwide inOxford New York

Auckland Bangkok Buenos Aires Cape Town ChennaiDar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata

Kuala Lumpur Madrid Melbourne Mexico City Mumbai NairobiSão Paulo Shanghai Taipei Tokyo Toronto

Oxford is a registered trade mark of Oxford University Pressin the UK and in certain other countries

Published in the United Statesby Oxford University Press Inc., New York

© Donald John Roberts 2004

The moral rights of the author have been assertedDatabase right Oxford University Press (maker)

First published 2004

First published in paperback 2007

All rights reserved. No part of this publication may be reproduced,stored in a retrieval system, or transmitted, in any form or by any means,

without the prior permission in writing of Oxford University Press,or as expressly permitted by law, or under terms agreed with the appropriate

reprographics rights organization. Enquiries concerning reproductionoutside the scope of the above should be sent to the Rights Department,

Oxford University Press, at the address above

You must not circulate this book in any other binding or coverand you must impose this same condition on any acquirer

British Library Cataloguing in Publication DataData available

Library of Congress Cataloging in Publication DataData available

Typeset by SPI Publisher Services, Pondicherry, IndiaPrinted in Great Britain

byClays Ltd., St. Ives plc

ISBN 978–0–19–829376–7 (Hbk.)978–0–19–829375–0 (Pbk.)

1 3 5 7 9 10 8 6 4 2

To Kathy

Contents

Preface ix

Preface to the Paperback Edition xiii

1. Strategy and Organization 1

2. Key Concepts for OrganizationDesign 32

3. The Nature and Purpose of theFirm 74

4. Motivation in the Modern Firm 118

5. Organizing for Performance 180

6. Organizing for Growth and Innovation 243

7. Creating the Modern Firm:Management and LeadershipChallenges 281

References 289

Index 303

vii

Preface

The most fundamental responsibilities of general man-agers are setting strategy and designing the organizationto implement it. Over the last decades the great value ofeconomics for the study of strategy and the practice ofstrategizing has become evident. This book attempts toshow how economics can contribute in a similar way andat a similar level to organizational design.

I hope that practicing managers will read and benefitfrom this book. It is not, however, a “how to” book offer-ing the final, simple answer about how to succeed.Instead, it offers ways to think about the problem ofdesigning business organizations for performance andgrowth. Both students of organizations and managementand practicing managers can benefit from having anunderstanding of the basic principles of the economics oforganization and its application to business enterprises.The book tries to offer this. It mixes case studies andshorter examples with fundamental conceptual and theo-retical material that is developed and presented in a non-technical manner and then applied to the design problem.It also seeks to explain some of the very great changes in

ix

actual companies that are creating the new model of themodern firm.

The lectures on which this volume is based were givenat Oxford in the spring of 1997, and it is the summer of2003 as I write this preface. Clearly, it has taken me a longtime to write up my lectures! Yet, I believe the delay wasprobably worthwhile. In the intervening time, there hasbeen much progress on the subject of how to put togethereffective organizations, and I personally have learned alot. Consequently, this volume is radically different thanit would have been if I had written it five or six years ago.In particular, three lectures have grown into seven chap-ters. There is new theory, and there are rich examples ofpractice that were not available then.

I owe much to many people. First, I am grateful for thehonor of having been invited to give the first ClarendonLectures in Management Studies, and I thank ColinMayer, the Oxford University School of ManagementStudies, and Oxford University Press. Second, almost allthe work I have done in organizations was collaborative,and I am indebted to each of the people with whom I havethought, taught, and written. I learned from all of them,but especially from Susan Athey, Jonathan Day, BengtHolmström, Paul Milgrom, and Joel Podolny. They willrecognize their ideas here and know how greatly I havebenefited from working with them. The StanfordGraduate School of Business provides an unmatched envi-ronment for teaching and research in organizations, and Iam grateful to the School for its support and to my facultycolleagues and the students in the Ph.D., MBA, Sloan, andExecutive programs for their huge contributions to mylearning. I am especially pleased to acknowledge my debts

Preface

x

to Bill Barnett, Dave Baron, Jim Baron, RobertBurgelman, Katherine Doornik, David Kreps, Ed Lazear,John McMillan, Charles O’Reilly, Paul Oyer, GarthSaloner, Scott Schaefer, Eric Van den Steen, and BobWilson. During the writing of this book I also spent timeat Nuffield College and at McKinsey & Company,London, and I want to express my gratitude to both thoseinstitutions and their members. I am also grateful to theexecutives and managers at the many companies I havebeen able to visit and study, especially BP, GeneralMotors, Johnson Controls, Nokia, Novo Nordisk, Sony,and Toyota. The cases I co-authored on these companieshelped to shape my thinking and are the basis for much ofthis book. My editor at Oxford University Press, DavidMusson, has shown immense patience with my tardiness(but not so much that I ceased feeling guilty!) and is owedthanks. Paul Coombes, John McMillan, AndyPostlewaite, Richard Saouma, and especially JonathanDay read the manuscript and offered useful comments.Ayca Kaya provided valuable research assistance and JenSmith was helpful putting the manuscript in final form.Finally, my wife, Kathleen Roberts, has suffered myinterminable dragging out of this project all these yearswith her usual grace and humor. Thank you, Kathy.

Preface

xi

xiii

Preface to the Paperback Edition

It is over three years since the initial publication of TheModern Firm and four years since I finished writing it. Inthat time, a number of significant developments haveoccurred in the study and practice of organizational design,particularly as informed by economics.

First, companies continue to develop the practice oforganizational design. The basic point of this book is thatit is the most fundamental responsibility of a general man-ager to craft a strategy and create an organization throughwhich that strategy can be successfully implemented inthe economic, political, legal, regulatory, social, and tech-nological environment in which the firm operates.However, the problem of finding alignment among theenvironment in which a firm operates, its strategy, and itsorganizational design is not one that can be solved onceand for all. Rather, it involves an on-going process ofadjustment as the environment changes, as the strategydevelops, and as the organization evolves. Certainly that isthe case with many of the firms I discuss in the book: theyhave adjusted their strategies and organizational designssince I first described them.

A prime example is BP, a company that features promi-nently in The Modern Firm. The discussion in the book isfocused on an especially innovative and effective organiza-tional design implemented in the early to mid 1990s. In thelate 1990s, however, BP made a number of very large acqui-sitions that doubled its employment, complicated its busi-ness and activity mix tremendously, and brought it into newand difficult geographies and political environments. Theclean, simple organizational model that had been so success-ful did not fit the new, complex environment and strategynearly so well. Consequently, BP has been experimentingwith adjustments to the model to regain alignment. Thefundamental principles on which BP operates have notchanged, and the tools that BP has employed in thinkingabout organizational design, many of which are laid out inthis book, remain in use. But BP’s executives and managersare working to adapt them to the company’s new scale andcomplexity and achieve effective coordination between thecompany’s center and its far-flung operations.

On the academic side, there have been important newideas emerging that will, I forecast, ultimately affect our basicunderstanding of the nature of the firm and of motivationproblems. Particularly exciting here are a set of ideas due toEric Van den Steen (see http://web.mit.edu/evds/www/)about the importance of differences of opinion among peo-ple in the company (about, for example, where is the bestplace to invest or what technologies will succeed) that arefundamental and not simply the result of the different peoplehaving different information. Van den Steen has used theseideas to explain formally why having a visionary manager –one who believes especially strongly in a particular vision ofthe future – can be motivating and rewarding for employees


xiv

http://web.mit.edu/evds/www/

and thereby good for investors. He has also used them toinvestigate the emergence of shared belief systems in organi-zations. And he has most recently employed them to give atotally new understanding of the nature of authority withinthe firm, why it is that the firm owns the tools with whichthe employees work, and why employees receive relativelymuted performance pay. These insights are fundamentallyimportant.

New evidence is also emerging. There is, as readers mayrightly complain, a dearth of systematic empirical researchby economists on organizations, and that is reflected in theimportance of case studies in the book. But recently a num-ber of scholars have been producing really very excitingresults. These involve immense amounts of work, becausethe data must typically be collected directly from individualfirms by the researchers, who cannot rely on trade associa-tions or governments to do it. Among the most striking is amajor project led by Nick Bloom (see http://wwwecon.stanford.edu/faculty/bloom.html) and John VanReenen (see http://cep.lse.ac.uk/people/bio.asp?id=1358).They have collected data on managerial practices (such asperformance evaluations, budgeting, and performancerewards) and organizational architecture (the number oflayers in the hierarchy, who makes what decisions) in thou-sands of firms across many countries and then related theseto performance. There is much still to do, but this work istremendously promising.

Both the new theory and the new evidence are reflectiveof the fact that more and more economists are being drawnto study organizations. Indeed, there is a legitimate sub-field of organizational economics that is emerging withinthe economics discipline, initially in business schools but


xv

http://www.econ.stanford.edu/faculty/bloom.html

http://www.econ.stanford.edu/faculty/bloom.html

http://cep.lse.ac.uk/people/bio.asp?id=1358

increasingly in traditional economics departments. A signalof this is that over forty distinguished economists are nowat work producing a Handbook of OrganizationalEconomics, which is meant to introduce the field to youngscholars. To be edited by Bob Gibbons and me, the volumeshould appear reasonably soon (by academic standards)!

At the end of 2004, the management editor of TheEconomist, not himself an economist by training,announced he thought The Modern Firm was the bestbusiness book of the year. This was of course an incredi-ble thrill for me: when I came across the review, late onenight, I rushed in and woke my wife, then breathlesslycalled friends in London because they would be awake,whereas my North American friends were all asleep, andI needed to share my joy. As gratifying as the recognitionwas, however, more important was the evidence it offeredthat there was a real chance that the book would be use-ful to practitioners, as I had hoped it would be. In fact, anumber of particularly insightful and generous execu-tives and consultants from several countries have let meknow they do indeed use the ideas from the book andhave invited me to meet with them and their colleagues.I thank them for their interest and support. Especiallystriking to me was that managers in the not-for-profitsector and in government have found these ideas useful.I also thank the canny publishers in some dozen differentcountries, from Estonia to Brazil and from China toFrance, who are producing translations of the book.

This paperback edition is meant to make my bookmore widely available. I seek your comments and welcomeyour questions. Please contact me at [email protected].


xvi

1

1

Strategy and Organization

During the first two decades of the twentieth century,managers at Standard Oil of New Jersey, Dupont, SearsRoebuck, and General Motors invented a new way of orga-nizing and managing their businesses. Their creation—thenow ubiquitous multidivisional form—involved funda-mental changes in the design of the firm. While the mostvisible change was structuring the organization on thebasis of divisions defined by product or geography, ratherthan functionally, the new form also involved new sys-tems for collecting and recording information, for allocat-ing resources, and for controlling behavior. This newmodel permitted an efficient solution to the incrediblycomplicated problem of coordinating and motivatinglarge numbers of people carrying out a complex of inter-related activities, often in different locations. It thusallowed giant, multiproduct business organizations toemerge and function effectively on a continental and thenglobal scale. The new design also led to a huge growth inthe number of people working as managers and to theemergence of the set of values and norms that mark man-agement as a profession. In terms of its impact, not just

on economic activity, but also on human life as a whole,the multidivisional organizational design must rank asone of the major innovations of the last century.1

Yet the last two decades have seen a set of innovationsin the organization of the firm that is similarly funda-mental and that may ultimately be as momentous. All theelements of the design are not yet in their final form.Managers continue to experiment with improving it asthey implement changes in their organizations. Still,some broad outlines are clear. Firms have changed thescope of their activities, typically refocusing on their corebusinesses and outsourcing many of the activities thatthey previously regarded as central. These changes arereflected in the immense volume of merger, acquisition,and spin-off activity that marked both the 1980s and1990s and that may now be building again. Many havealso redefined the nature of their relationships withcustomers and suppliers, often replacing simple arm’slength dealings with long-term partnerships. They haveeliminated layers of management and associated staffpositions, redefined the units into which they dividethemselves internally, dispersed functional experts to thebusiness units, and increased the authority and account-ability of line managers. By these measures, coupled withimproved information and measurement systems andredesigned performance management systems, they havesought to increase the speed of decision-making and totap the knowledge and energy of their employees in waysthat have not been tried before. To facilitate coordinationand learning, they have experimented with linkingpeople in different parts of their organizations directly,so that communications are more horizontal and not


2

just up and down the hierarchy. Many have also tried toredefine the nature of the relationship they have with theiremployees while redesigning jobs and the very nature ofwork.

These changes are aimed at improving the performanceof the firms adopting them. Increased competitive pres-sures drive their adoption, and new technology makesmany of them feasible for the first time. Falling barriers tointernational trade and investment, the rise of informa-tion technology (especially the Internet), and improvedtransportation mean that a firm’s competitors are not justthe old local rivals, but may be from anywhere. With morecompetition, the need to improve performance increases.These same developments also open new opportunities todo business far from home, and the new organizationaldesigns support taking advantage of these opportunities.Capital markets, too, are increasing the performance pres-sures on firms. Especially in the United States, butincreasingly elsewhere as well, the increased power ofinstitutional investors and their increased willingness touse this power are pushing companies to do better. Insome cases, the changes are also responses to greater com-petition for talent as more firms seek to attract and retainespecially skilled and gifted people. Meanwhile, the mas-sive advances in the technology of communication andcomputation make feasible many of the key changes inorganization and management that are being adopted.

These organizational innovations, when properlyapplied, do lead to better economic performance, affect-ing the material well-being of the people of the world.Moreover, they alter the ways work is done, changingpeople’s lives in fundamental ways. Ultimately, they can


3

affect every aspect of how business is accomplished in themodern firm.

Many of the principles underlying this new model arenot in fact completely new, however, as the followingexample will show.

The example involves two firms in a service industry—trade. One, the long-established “HB Company,” hadcompletely dominated the market for years. Its leaderswere politically as well as economically powerful, and thefirm was favored by successive governments. The newlyestablished “NW Company,” the upstart rival, had noneof these advantages. Its leaders were immigrants andrefugees, its headquarters was in a distant, provincialtown, and it had no powerful friends. In fact, the NW Co.was arguably breaking the law in even attempting to com-pete with the HB Co. Further, in addition to its advan-tages in an established customer base, in businessexperience, and in political and legal matters, HB Co. hadvastly superior technology and better access to financing.The result was that HB’s costs were estimated to be inthe order of one-half those of its rival.

Yet, in a relatively brief time after entering the busi-ness, NW Co. had seized 80 percent of the market fromits rival and was very profitable, while the once-dominantmonopolist was near bankruptcy. How did this happen?

The answer will be unsurprising to anyone familiarwith the changes that have gone on in business recently.The NW Co. found a way to serve the customers betterby getting closer to them, where it could be more respon-sive to their differing needs and to ever-changing marketrealities. It also carried out a number of organizationalinnovations. It simplified the supplier structure and


4

eliminated traditional middlemen. It avoided excessivebureaucracy while developing systems to ensure thatrelevant information was shared broadly within thecompany and that all the relevant parties had a role in andunderstood decisions. It recruited people for operatingpositions who were willing to take responsibility and ini-tiative, and then gave them the authority to act on theirknowledge and intelligence without checking every detailwith hierarchic superiors. Finally, it put in place rewardsystems that encouraged entrepreneurial behavior. Inother words, it developed a new strategy and then put inplace the people, organizational structure, managerialprocesses, and corporate culture to support the strategy.These managerial innovations allowed it to overcome anapparently prohibitive cost disadvantage.

HB Co. was initially unperturbed by the challenge fromNW. It knew that its ways had worked for years and that ithad tremendous advantages. It probably also failed to see thecompetitive advantage that NW’s new strategy and organ-ization provided. So its response was very slow in coming.Even after the upstart rival had gained a huge market share,the leaders of the old firm did little. This, too, should be afamiliar story to those who have followed the experience ina number of different businesses in recent decades.

Eventually, however, HB did respond to the threat,essentially by copying NW’s new approach. It did so,however, only after the leaders of the firm had beenreplaced by new ones who understood the nature of thethreat and who were not tied to the old ways that hadworked so well for so long.

NW Co. had always known that it would be doomed byits cost disadvantage if HB Co. were to copy its rival’s


5

customer-oriented strategy and replace its centralized,command-and-control processes with ones that fit thenew strategy. So NW attempted a preemptive takeover ofHB before the new leadership could take control. Thisfailed, however, and ultimately HB Co.’s huge costadvantage did overwhelm the NW Co.

The outcome was a deal uniting the two firms that, forpublic relations reasons, was labeled a merger. But every-one in Canada in 1820 (and those in the United Kingdomwho were aware of the matter) knew that the North WestCompany of Montreal had been absorbed by the victori-ous, London-based Hudson Bay Company, or moreproperly, the Governor and Company of Adventurers ofEngland Trading into Hudson’s Bay.2

The Hudson Bay Company continues in business tothis day as one of the leading retailers in Canada.Established by Charles the Second’s royal charter in 1670under the leadership of Prince Rupert, the HBC had beengiven exclusive rights to trade in the lands draining intothe giant Hudson Bay. This monopoly grant covered anarea of 1.5 million square miles, more than fifteen timesthe size of the United Kingdom and significantly largerthan the European Union before its expansion in 2004. Atthat time there were no Europeans resident in this area,which was a trackless wilderness of rocks and trees andwater (as much of it still is today!). What it did containwas a relatively small number of aboriginal people anduntold amounts of animal furs, especially beaver, whichwere in high demand in Europe.

The Company exploited its franchise by a very passivestrategy: It built half a dozen forts on the shores of thebay and waited for potential customers to come to it,


6

seeking European-made goods for their furs. The tradegoods were brought in from England through HudsonBay in ocean-going ships that made annual voyages,bringing the furs back to England on their return trips.(More frequent trips were impossible because of the tech-nology of shipping and the fact that the bay was frozensolid for most of the year.) The HBC stuck to thisapproach over the next century, during which a trade net-work developed in which aboriginal tribes, whose home-lands were located away from Hudson Bay, traded withones nearer the forts, who then traded with the HBC.

This approach to business was hardly very bold, but itwas a sensible strategy, given the market conditions andtechnology of the seventeenth and eighteenth centuriesand the risks and opportunities they implied. Moreover,the HBC built an organizational system that fit thestrategy very well and that allowed it to be implementedvery effectively.

Key decisions were centralized in London. This meantthat decision-making was slow and unresponsive to localconditions (especially since none of the senior decision-makers ever set foot in the company’s territory, calledRupertsland), but it did ensure coherence and control.Moreover, with an unchallenged legal monopoly, a pas-sive approach to business development, and an unsophis-ticated, slowly changing market, there was little obviousneed for speedy decisions. Rather, the danger was thatlocal employees, far from the oversight of senior manage-ment, would fritter away the profits, or, worse yet, misap-propriate them. So the Bay’s people in Rupertsland wereselected as much for their lack of imagination and theirability to bear tedium as for their talent, initiative, and


7

diligence. They were sent out under contracts that wereclose to indentured servitude, given a specific set ofdetailed instructions that governed every aspect of theirwork (including the allowable prices to pay and charge),paid a fixed amount, required to stay near the company’sforts, and punished physically for any infractions.

The system seems brutal and stupid, and yet it mustbe recognized that it worked very well: The Companywas very profitable from the outset and remained sothroughout its first century of existence.

There was, however, an inherent inefficiency in thissystem. It did not do a good job of exploiting the oppor-tunities for trade with people far from the bay, leavingtheir (perhaps unimagined) desires for European goodsunmet and the furs they collected in hunting for foodunderused. Indirect trade through middlemen from thepeoples living between the Company’s forts and the richfur areas allowed partial realization of these potential gainsfrom trade, but the system was arguably inefficient. Thefirst reason to expect inefficiency is that the middlemenhad monopoly positions that they likely exploited, so thatmarkups were taken on markups and the volumes trans-acted were too low. The second is that the middlemenwere poorly positioned to bear the risk associated with thetrade. They lacked access to finance to support their marketpositions and had to face the uncertainties of demand andsupply on their own. Both these effects limited the actualvolume of trade to inefficiently low levels.

The founders of the North West Company—recentimmigrants to Montreal, either directly from Britain or asrefugees from the revolution in the thirteen Americancolonies—may have seen the profit opportunities that


8

were inherent in these inefficiencies. They probably wereaware that French Canadian traders had profitably tradedmore directly with the native people before the Britishconquest of Canada, and they certainly saw the profitsthat the Hudson Bay Company had been racking up formore than a century despite its passive strategy. Adecently effective competitor ought to have been able todo very well.

The Nor’Westers’ big disadvantage, however, was thatthey could not ship directly in and out of the center ofthe fur-trading area: Hudson Bay was closed to them bythe HBC’s monopoly grant. Instead, they would have tobring the trade goods in from Europe and the furs backout through Montreal, the head of navigation on theSt. Lawrence River. But Montreal was thousands of milesfrom the areas where furs were richest and most plentiful,and, in particular, almost a thousand miles further fromthem than were the HBC forts. The Nor’Westers couldnot expect the potential customers to come to them, andso they were forced to go to their customers.

This was the origin of their strategy, and the source oftheir huge cost disadvantage. In the last decades of theeighteenth century the Nor’Westers set up dozens oftrading posts right in the lands where the furs were col-lected, reaching all the way to the Athabaska region inwhat is now the far north of Saskatchewan and Alberta.Then, in birch-bark canoes and small open boats paddledby French-Canadian voyageurs, they brought the tradegoods into the wilderness and the furs out to market,from Montreal to the Athabaska and back, through theGreat Lakes and along the untamed rivers of theCanadian North.


9

This strategy, operating in a now-competitiveenvironment, required very different structures, proce-dures, and behavior to make it work from those the HBChad used for 120 years. Given the communicationstechnology of the period, coordinating such a complexoperation could not be done through central decision-making. Instead, individuals in the field would need to takeresponsibility for coping with unforeseen eventualities andchanging conditions as they arose. To ensure that this wasdone effectively, the men who actually ran the posts in thefur country were partners in the company (“winteringpartners”) with broad authority over their operationsbacked by the incentives of ownership to do a good job.Also, getting the trade goods in and the furs out was a mon-umental task that relied on near superhuman physicalefforts. The wintering partners had the direct incentive ofa profit share to ensure that these tasks were fulfilled andthe efforts exerted, and they in turn gave strong incentivesto the voyageurs to carry out the work (including the pos-sibility of becoming a partner). Meanwhile, the Montreal-based partners handled the acquisition of trade goods, saleof the furs, and financing the operations. They also handledgetting the trade goods to the company’s inland headquar-ters at the head of Lake Superior, where they met eachsummer with the wintering partners, who had brought outthe furs from the North. This annual meeting of all thepartners ensured that information was shared and thatdecisions were informed and understood.

What are the lessons of this example, other than, per-haps, that there is nothing new under the sun?

First, strategy and organization matter: The NorthWest Company’s strategy of eliminating middlemen and


10

getting close to the customer, backed by an organizationthat implemented this, quickly overcame a 50 percent costdisadvantage, over a hundred years’ experience, and aroyal grant of monopoly.

Second, there needs to be a fit between strategy andorganization and between these and the technological,legal, and competitive environment. The organization ofthe HBC fit its strategy and the environment thatobtained until the entry of the North West Company, andthe result was a century of profitability. The Nor’Westers’model similarly showed an internal coherence and analignment with the strategic, technological, and competi-tive context. In general, however, finding such a fit wouldseem a daunting challenge, because there are so manyvariables and the choice is so complex. Still, it can bedone, and it must be if success is to be achieved.

Third, strategic and organizational change is not easy,but it is sometimes necessary and it can and does happen.The HBC took a decade to reform in response to thethreat. That is almost as long as it took the Americanautomobile industry to respond to the successful entry oftheir Japanese rivals! Finally the changes came, althoughonly under the threat of bankruptcy. The HBC put trad-ing posts inland to meet the competition, reformed itsorganizational processes to support the new strategy, andultimately triumphed.

Fourth, a more competitive environment favors thesort of organizational design that the North WestCompany initiated and whose principles are shared in theemergent organizational design of the modern firm.

In this book I will seek to elucidate these principles andshow how they apply. In the process I will develop some


11

conceptual frameworks and theoretical constructs thatare important for understanding effective organizationaldesign.

My starting point for this exercise is the proposition thatgeneral managers must be organizational designers. Just asit is a fundamental responsibility of general managers todevise a strategy that determines how their businesses willcompete, it is equally necessary that they design and createan organization through which the strategy will be imple-mented. And just as we have come to realize that strategyis not solely the responsibility of the chief executive officer,but rather of managers throughout the organization, so toois organizational design.

The second basis for this book is the idea that econom-ics has much to say about the problem of organizationaldesign. In the twenty-plus years since Michael Porterbegan applying the concepts of industrial organizationeconomics to the field of strategy (Porter 1980, 1985),practitioners and students of management alike havecome to recognize that economic analysis is of tremen-dous value to this field. The methods of economics holdsimilar promise for the study and design of organizations,as I hope the following will demonstrate. But first weneed to set some context.

Strategy, Organization, and the Environment

Achieving high performance in a business results fromestablishing and maintaining a fit among three elements:The strategy of the firm, its organizational design, and theenvironment in which it operates. In the conceptualization


12

that has been standard in management studies, theorganizational design problem takes the economic, legal,social, and technological environment in which the firmoperates as given, presumes that the strategy has been for-mulated, and then seeks to create an organization to imple-ment the given strategy in the particular environment.This approach follows from Alfred Chandler’s dictum(Chandler 1962) that “structure follows strategy”—organ-ization is the mechanism through which strategy is realized.

Although this is a very limited view of the nature of thedesign problem and of the role of organization, we willplace our discussion initially in this context. For the sakeof simplicity, then, focus on the traditional Chandlerianformulation and consider a simplified, idealized versionof what a firm is and does. The starting point is a businessopportunity—an unmet need, a market inefficiency. Forthe NWC, the opportunity lay in the HBC’s inefficientexploitation of the potential gains from trade. Moregenerally, the opportunity might come from having lowercosts than the current market participants or a productthat better meets the needs of (at least some) customers.This, in turn, might reflect better technology, or morecreativity, or previously unexploited economies of scaleand scope.

Next, in the traditional view, comes a strategy to exploitthat opportunity—a specification of how the firm is going tocreate value and get to keep some of it. A well-formulatedstrategy has several components (Saloner, Shepard, andPodolny 2001).

First, a strategy involves a goal against which the firmcan measure itself and judge its success. This might be pro-fit or shareholder-value maximization, or it might be


13

something more complex involving the interests of differentconstituencies and stakeholders. Even when shareholdervalue is the ultimate objective, the strategic goal might beexpressed in more operational (and more motivating)terms. For example, in the 1970s and 1980s, Komatsu, theJapanese heavy equipment manufacturer, had as its aim to“Beat Caterpillar!”

The next key element is a statement of scope—aspecification of the business the firm is in, what productsand services it will offer, what customers and marketsegments it will serve, what activities it will undertake,where it will do these things, and what technology it willuse. Obviously, the choice of what to do and how, where,and for whom to do it is a directly relevant and significantaspect of strategy. Less obviously, the scope of the strat-egy determines what opportunities the firm is not goingto pursue. This is important: Strategy is a disciplinedevice that helps sort out which of the myriad opportuni-ties that will arise the firm should pursue and on which itshould pass. It also allows people in the organization tomake this determination without a lot of further discus-sion and debate, so it facilitates coordination. Moreover, itcan contribute to motivation by providing clear goals andboundaries for choices.

A third key element of a strategy is a specification ofthe nature of the firm’s competitive advantage, an indica-tion of how the firm’s offer will lead others to deal with iton terms that allow it to realize its goals. How will itattract a profitable market? How will it create value, gen-erating a willingness to pay by customers that exceeds thecosts of serving them? Will the firm offer a better productat a cost increment that is lower than the additional value


14

of the improvement to customers? Will it offer as good aproduct for less? A less desirable product but for a muchlower cost?

The final component of a strategy is an explication ofwhy the claimed competitive advantage will actually berealized. Why will the firm get to claim some nontrivialshare of the value it creates and do so in a sustainable way?How will the firm get a price that exceeds its costs? Whatwill keep actual and potential competitors from erodingits margins and stealing away its customers? What willensure that suppliers or customers do not manage toappropriate all the value created? This piece is often miss-ing in formal strategy statements, but the existence andvalidity of such a logic are crucial. Typically, a valid logicwill involve a system of implications linking the particu-lar position occupied by the firm and the distinctivecapabilities it enjoys to the customers’ choices and thenback, via the prices, costs, and volumes that result, to thefirm’s ability to maintain and enhance its position andcapabilities.

Had the NWC’s leaders enunciated their strategy itwould thus have been something like the following:

The NWC will trade with the native people of the Canadiannorth, taking furs in return for European goods. The tradewill occur at posts established in the fur-bearing regions andthe transport between the posts and Montreal will be pro-vided by company employees using small watercraft. Thetrade goods will be obtained in Montreal and England, andthe furs will be sold in London. The NWC will offer termsof trade that are better than the effective net ones comingfrom the HBC through the middlemen who are between itand the people actually collecting the furs. The NWC will


15

also be more responsive to the customers’ needs than is thecompetition. Together, these will make it the preferredtrading partner. This positional advantage and the savingsfrom eliminating the middlemen will allow the NWC toserve its customers on terms that still leave a profit margin,despite the Company’s higher costs. It will be able to offersuch terms and keep these profits as long as the HBC doesnot match its offer [which the established firm’s strategy,organization, and management initially prevented it fromdoing]. This will allow the NWC to achieve its goal of prof-itably dominating the fur trade in British North America.

In a multi-business firm, there is another level to strat-egy, that of corporate strategy. A corporate strategy iden-tifies the set of businesses the firm will encompass and thelogic of why doing so will allow it to create extra value overand above what a collection of stand-alone businesses cancreate. Thus, it is essentially a portfolio choice combinedwith a theory of the role of the corporate center.

A strategy implies a set of activities that need to becarried out to realize it. In a typical firm these include the“value-chain” activities that must be undertaken in meetingcustomers’ needs, such as product design and development,input procurement, manufacturing, distribution, sales, andpost-sales service, as well as “support” activities such ashuman resource management, management informationsystems, and finance. In the NWC the value chain wasacquisition of trade goods, their transport to the customers,the actual trade, and transport of the furs out to Montrealand thence to London for sale.

The organization is then the means through whichthese activities are to be carried out and the strategy is tobe implemented. Any firm’s organization is multifaceted,


16

and the range of organizational variables is mind-boggling.Thus, even a bit of classification may be of some use. Onetaxonomy identifies the organization as a collection ofpeople and an array of organizational features. These, inturn, can be sorted into architecture, routines, and culture,giving rise to the acronym “PARC.”

First is the set of people who are part of the organiza-tion. What sort of talents and skills do they have, whattastes, what beliefs, what objectives? How hard are theyprepared to work and for what ends? What sorts of riskswill they accept and what sorts of rewards do theyvalue? How are they connected to the firm? As owners?Employees? Contractors?

The architectural features include what is on theorganization chart: the vertical and horizontal boundariesof the firm; the assembling of tasks into jobs and jobs intodepartments, business units, and divisions; the reportingand authority relationships; and so on. It also includessuch matters as the financing, ownership, and governancestructure of the firm. These are relatively “hard” features,often with an explicit contractual element. However,architecture also includes the personal networks that linkpeople throughout the firm and across the firm’s bound-aries. These can, in fact, be as important and more thanthe formal architecture.

The routines include all the managerial processes,policies, and procedures, official and unofficial, formal andinformal, that shape how information is gathered andtransmitted, decisions made, resources allocated, perfor-mance monitored, and activities controlled and rewarded.The allocation of decision authority within the firm—whatdecisions are made by which people at what levels, with


17

what oversight or review—is a key element here. Theprocesses also include the routines through which work isdone and the mechanisms through which these are altered.These features may involve explicit contractual elements aswell as “implicit contracts,” more or less formal, sharedunderstandings about how things are to be done.

Culture is the “softer” stuff, but it is not less impor-tant for that. It involves the fundamental shared valuesof the people in the firm, as well as their shared beliefsabout why the firm exists, about what they are collec-tively and individually doing, and to what end. It alsoencompasses the special language used within the firm,which shapes thought and action. Culture also involvesthe fundamental mindsets of the firm’s members andthe mental models they have, which determine how theysee themselves and the firm and how they interpretevents. Most significantly, it involves the norms ofbehavior that prevail in dealing with other members ofthe firm and with outsiders. Culture defines the contextin which the relations among people develop and oper-ate and sets the basis for the implicit contracts thatguide and shape decisions. It operates as a social moti-vation and control system.

Along with the strategy and the organization, thethird determinant of performance is the environment inwhich the firm operates. This includes its competitorsand their strategies and organizational designs, the stateof other relevant markets and firms (suppliers of inputs,complements, and substitutes), and the customers, as wellas the ambient technology, the legal and regulatory con-text, various political, social, and demographic features,and so on.


18

The Design Problem: Setting Strategy andOrganization

If we now apply a design perspective, the job of thegeneral manager is to craft a strategy—objective, scope,competitive advantage, and logic—and create an organi-zation—people, architecture, routines, and culture—inlight of the environment to maximize performance. In thelonger term, the designer might also try to shape theenvironment, but we will largely leave this aside in ourdiscussion. The model is captured in Figure 1.

Performance thus depends on the strategy, the organ-ization, and the environment. This formulation leads to acontingent theory of strategy and organization. There isno uniquely best strategy, and there is no one best way toorganize. The attractiveness of a strategy is defined onlyin terms of how well it works in the environment inwhich it is operating with the organization that is tryingto implement it. Similarly, the value of an organizational


19

Designer

Strategy Activities Organization

Environment Performance

Figure 1. The design problem is to select the strategy andorganization to achieve maximal performance in the context ofthe environment

design depends completely on how well it matches theparticular environment and strategy. What is good is whatworks, and we should expect that what will work in dif-ferent contexts may depend on the context. The key is infinding and establishing a fit among strategy, organiza-tion, and the environment and then maintaining thefit over time in the face of change.

What, however, is “performance?”Firms are institutions created to serve human needs.

Performance ultimately is how well the firm does at serv-ing these needs. This raises the issue of whose needs areto be served. Is the firm simply a mechanism for creatingshareholder returns? Or is it (also) to provide meaningfulexperiences, secure employment, and valued opportuni-ties for its members; valuable goods and services for cus-tomers; tax revenues and jobs for communities; positiveeffects on the environment; and so on?

We will, for the most part, take the point of view thatthe purpose of the firm can be expressed as “value cre-ation.” This is not an uncontroversial position, either onprescriptive or descriptive grounds. Indeed, it should noteven be immediately clear what it means. The value cre-ated by economic activity is the difference between themaximum that people would be willing to pay for it, lessthe opportunity costs of the activity. Under rather spe-cific and somewhat special conditions, value and valuemaximization are well defined and would be uncontrover-sial objectives. These conditions are that (1) there is amedium of exchange that is valued by everyone, (2) thatthis “money” is freely transferable in any amountbetween people, and (3) that the amount of money thatjust compensates any individual for any change in his or


20

her circumstances does not depend on how much moneythe individual already holds. Under these conditions, thevalue created in any act of production or trade is just theextra amount of money that all parties in aggregate wouldpay (or need to be paid) in order to induce unanimousagreement for the action from all affected parties. Thisamount is unambiguously defined and, further, maximi-zation of the value created in this sense is equivalent toachieving full economic efficiency. If the value-maximizingcourse of action is adopted, it will be impossible to findany alternative that all parties would unanimously prefer,and if there is another course of action that creates greatervalue, then it will be possible to make everyone better offby adopting this alternative and distributing the gainsappropriately.3

Under these conditions, then, value maximization isarguably an appropriate goal from a social point of view.Further, to the extent that those in whose interests thefirm actually operates are able to claim the value created,they would want it to be run so as to maximize value. Ofcourse, the conditions are restrictive, and they surely arenot fully met in the real world. Even if we treat financialwealth as the universally desired good (as seems most nat-ural), the second condition may fail if the winners fromsome move do not have enough money to compensate thelosers, and in this case the condition of maximizing valuemay not win unanimous support. It is also necessary thatall the relevant interests are recognized and taken intoaccount. Moreover, the third condition requires that therebe no “income effects” in demand, which is surely false.

Nonetheless, with a reasonably complete, well func-tioning system of markets and contracts, the assumption


21

that the owners of firms want their companies run tomaximize the owners’ long-term wealth is likely to bedescriptively accurate (at least to a first approximation).Further, if there are adequate mechanisms for taking careof concerns that are not reflected in market prices, valuemaximization is unlikely to be an obviously undesirableobjective from a social point of view. For example, effec-tive markets will mean that employees have good outsideopportunities and so owners will not find it worthwhile toexploit the workers, and effective contracting will ensuretheir ability to protect themselves and perhaps claim ashare of the value created. Meanwhile, the effective lawsand regulations will lead the owners to want the firm notto abuse the environment or collude with competitorsagainst the customers’ interests.

There is still a problem with measuring long-termvalue or wealth creation. If stock market prices immedi-ately and accurately reflected all the available informationabout the firm and its prospects, then stock marketvaluations would be a fine measure, and maximizing themarket value of the firm would be an appropriate goal formanagers. Of course, markets do not necessarily work thiswell, especially in the short run. Moreover, if informationis deliberately withheld or manipulated, they cannot workwell. Still, over the long haul, honest managers who pur-sue the maximization of firm value are likely acting to cre-ate the most possible value for their shareholders. Thenthe problem becomes one of selecting the (long-run)value-maximizing strategy for the particular environmentand then creating the organization that will best realize it.This problem of organizational design is the subject ofthis book.


22

Strategic and Organizational Change

The foregoing formalization of the process of establishing afit between the environment, the strategy, and the organiza-tion is clearly appropriate for a start-up enterprise decidingfor the first time what it is going to do and how it is going todo it. The design approach applies as well in an ongoingfirm, although some interesting complexity emerges.

At any point in time for an established firm there willbe an issue of whether the existing strategy and organiza-tion generate the highest performance available in theenvironment in which the firm finds itself. Since theenvironment is changing, there is a likelihood that whatmay once have been a good fit has since deteriorated.This means that there can be a need for strategic andorganizational change.

Strategy can be changed relatively quickly: In principlea new strategy can be developed and announced in a shorttime. Organizations, however, show a lot of inertia, in twodistinct senses. First, successful organizations tend to per-sist, becoming long-lived assets in which firm’s strategiccapabilities are embedded. Thus, the existing organiza-tion shapes the opportunities for future strategic choiceand for responding to environmental change. Second,organizations cannot be changed as surely and quickly ascan strategy. While it is easy enough to change the formalarchitecture, it certainly takes real time to change the setof people in the firm and the networks among them, toredefine the fundamental beliefs they share, and to inducenew behavioral norms. Yet these may be the mostimportant elements to the realization of the strategy.Thus, effective implementation may not be immediately


23

possible. This will affect what strategic choices are best ina particular context.

The interesting complexities arise from the interplaybetween the speeds of change of strategy, organization,and the environment. In the traditional view of manage-ment scholars, the environment has often been taken tobe relatively stable, changing only slowly and infre-quently. This may well have reflected reality in an earlierday—it was certainly true in the case of the HBC in thefirst century of its history—and it may still apply insome industries today. If it is an approximately accurateassumption, then the approach already outlined appliesmore or less as indicated. Once the environmental changehas occurred, the new environment can be taken as givenand likely to persist. A strategy can then be developedto address the new opportunities the environment offers,and an organization put together to implement the strat-egy in the new environment, just as the traditional recipedirects. Should the environment happen to change again,the strategy can be altered and all aspects of the organiza-tion restructured to fit the new environment. Becausesome aspects of the organization will not adapt instanta-neously, there may be some period of misalignment. Sucha period would also result if the right organizationaldesign is not immediately obvious, but has to be discov-ered through a process of search and experimentation.However, in either case the misalignment ought to bebrief compared to the period over which the strategy andthe implied organization are in place and functioning.

BP Exploration (BPX), the “upstream” part of thecompany then called British Petroleum that was responsiblefor finding and producing crude oil and natural gas,


24

operated essentially according to this recipe in the late 1980sand first half of the 1990s (Berzins, Podolny, and Roberts1998a, b). When John Browne took over as head of BPX in1989, “petropreneurs”—smaller energy firms operating ina relatively narrow range of activities—often outperformedthe major integrated oil companies. Browne reformulatedthe strategy at BPX to seek cost advantage by focusing onwhat the oil industry called “elephants”—very large hydro-carbon deposits—that offered possible economies of scaleand took advantage of BP’s technical capabilities and finan-cial strength. This led to a redirecting of exploratory activ-ities and the sale of a variety of productive assets that did notmeet the test of being sufficiently large. This strategyremained in place throughout the succeeding years.

Meanwhile, at the corporate level, BP had beenorganized as a complex matrix, with geographic andstream (exploration and production, refining and market-ing, and chemicals) dimensions, large central staffs, andheavy centralization of authority. A shift in corporate orga-nizational design beginning in 1990 then moved signifi-cant decision power from headquarters to the streams.Beginning in 1992, in the context of a corporate financialcrisis that saw BP on the verge of bankruptcy, Browne inturn radically redesigned his part of the business. Heeliminated the regional structure in the upstream, as wellas most of the managerial center that set direction andoversaw operations. Decision rights were reallocatedbetween an extremely lean Executive Committee consist-ing of Browne and two other senior leaders and the man-agers of individual production units or “assets.” (Thetypical asset was a single oil field.) The asset managerswere empowered to determine how they would deliver


25

performance against negotiated contracts, which initiallyspecified targets for costs, capital expenditures, and vol-umes, and rewards were linked strongly to the perfor-mance of the individual assets. Meanwhile, the centralizedfunctional staffs were largely dispersed to the assets. Theobjective was to increase performance by empoweringthose closest to the relevant information to act upon it andmotivating them to do so, and the size of the individualassets meant that there was significant payoff from effortsaimed at improving performance at the asset level.

Employing this basic organization design resulted in agreat improvement in performance from the outset. Still,over the next five years, while the strategy remainedunchanged, there were important adjustments in the orga-nization on an ongoing basis as the company experimentedwith improving the design. Some of these changes wereexplicitly intended to overcome shortcomings in the orig-inal model. Chief among these was the introduction of“peer groups” that linked assets facing similar technicaland commercial challenges to provide mutual support andspread learning. This structure proved necessary becausethe absence of middle managers and the limited functionalexpertise at stream headquarters meant that the assetmanagers could not look to the center for help with prob-lems. Other changes were made in the organizationalarchitecture and routines to take advantage of the evolvingculture. Norms of trust, of helping other businesses, andof delivering promised performance became thoroughlyembedded in BPX under Browne’s leadership. Theseshaped behavior and thus allowed shifting the basis ofcompensation to promote other objectives than simpleindividual asset performance. They also permitted passing


26

responsibility to the peer groups for allocating capital andfor meeting the aggregate of the performance contracts ofthe member assets. These changes, in turn, permittedfurther performance improvements.

Strategy and Organization in TurbulentEnvironments

When environmental change becomes extremely rapidand ongoing, however, the sequential approach to strategyand organization exemplified by BPX may no longer beviable. An extended process of adaptation is not going towork—by the time a new strategy is generated and theorganization restructured, the environment will havechanged many times again.

In fact, many management scholars and practitionershave asserted in recent years that, in sufficiently turbulentenvironments, ex ante strategizing from the top becomesnearly pointless. The necessary information about marketsand technologies is not directly available to top execu-tives and it cannot be communicated to them and com-prehended with sufficient speed and clarity to be used fortop-down strategy formulation.

This position tends to confuse detailed, short-termtactics and formalized strategic plans (of the sort that getbound in fancy covers and then set on shelves, never to beread) with strategic thinking of the sort embodied in astrategy statement. Still, to the extent that the argumenthas some validity, then the nature of the design problemis altered in another interesting fashion.

In very turbulent environments, many of the specificsof the firm’s strategy are likely to emerge from a multitude


27

of decisions taken at various levels within the organization.The most that can be done from the top is the setting ofbroad strategic direction or intent. The design of theorganization then determines in large measure what deci-sions will get made. Thus, reversing Chandler, strategyfollows organization.

The solution, then, is to try to set the relatively slow-moving elements of the organization appropriately and thento take them more or less as given. What is appropriate inthis context clearly involves some forecast of the evolutionof the environment and of the basic strategic direction.Then the idea is to alter the strategy and some of the moreformal but malleable elements of the organization to keepup with the changes in technology and markets.

Yet the design perspective remains valid, even in thiscontext. Now, however, the role of the designer is to shapethe relatively inert elements of the organization, like theculture, that will exert a persistent effect on the strategicand organizational choices that are made by the people inthe firm, and to design a set of processes that will allowthem to make good decisions. The designer should alsoset broad strategic intent to inform and shape the dispersedstrategic decision-making. Finally, the designer mustadapt the strategic intent and the controlled elements ofthe organization over time.

Nokia Corporation, the Finnish manufacturer ofmobile telephones and network equipment, followed thislatter model during the 1990s (Doornik and Roberts2001). This was a decade that saw massive environmentalchange in the mobile telephony business: Deregulationand privatization, the entry of new service providers, anunforeseen explosion of demand, the emergence of digital


28

technologies, and, at the end of the decade, the beginningsof convergence between the Internet and mobile phones.Nokia was the big winner in all this—it went from nearlybankrupt in 1992 to being Europe’s most valuable companyin 2000.

The top executives of Nokia set the broad intent thatgave the context for the hundreds of dispersed decisionsthat became the firm’s strategy. In 1992 the intent wasexpressed in four criteria—“Focused,” “global,”“telecommunications-oriented,” and “high value-added”—under a vision that “voice will go wireless.” Thisled the company to exit the broad range of businesses thathad contributed 90 percent of its revenues only a few yearspreviously and to focus on mobile handsets and networkequipment. The leaders also set a stretch goal of doublingNokia’s market share by the end of the decade. When theobjectives for 2000 had already been realized in 1997, theearlier strategic intent was replaced by the statement thatNokia wanted to become the leader in the most attractivetelecom segments. (At the time, Motorola was the clearindustry leader.) By 1999, Nokia had indeed become theindustry leader in mobile phones, while the possibilitiesfor the Internet being accessed via mobile phones werebecoming evident. The 1997 objective in turn wasreplaced by the stated intent for Nokia to lead the devel-opment of the “mobile information society” by being thecompany that would “bring the Internet to everyone’spocket.” Note that these are not strategies. While they doaddress the scope issue broadly, they are not very preciseand they give little of the logic of why the firm isgoing to be able to create value or keep some of it. They do,however, set the context for strategy to emerge.


29

Meanwhile, the leadership of Nokia was very consciousof the importance of culture in motivating people to act inthe needed ways. It thus worked hard to keep the crucialfeatures of the culture fresh and effective, even as the com-pany grew at 30 percent per annum and spread its activitiesaround the world. As for the more formal elements of orga-nization, these were kept very fluid: “We hate organizationcharts, and if forced to create one, we will draw it only inpencil!” Formal structure was shifted almost constantly tokeep up with emerging needs and networks were used heav-ily to share knowledge and get work done. At the same time,the increasing size and complexity of the firm were met byan increased reliance on regularized processes in place ofearlier informal routines that had differed across the firm.

BP and Nokia were both extremely successful in the1990s, and they continue so today. Each adopted a distinctorganizational model that was well suited to the environ-ment in which the company operated and the strategy itpursued. The resulting fit between environment, strategy,and organization was key to the two firms’ successes: Likethe HBC and NWC centuries ago, and the firms thatdeveloped the multidivisional form at the start of the cen-tury, each solved the organizational design problem, andsuccess followed.

Notes

1. The standard reference on these developments is Chandler(1977). See also Chandler (1962).

2. The discussion here draws heavily on Newman (1985,1987). See also Newman (1991) for the history of the


30

company after the merger, and Spraakman (2002) for amore detailed discussion of aspects of the control systems ateach company, before and after the merger.

3. For more on the concept of value and its applicability, seeMilgrom and Roberts (1992: 35–9) and Roberts (1998).


31

32

2

Key Concepts for OrganizationDesign

The problem of strategic choice and organizational designis, in principle, immensely complex. Selecting a strategy iscomplex already, but when one considers all the elementsof an organization design, the problem becomes mind-bogglingly complicated. Indeed, the “rugged landscapes”literature in strategic management (Levinthal 1997) isbuilt around the idea that the problem is arbitrarily com-plicated, essentially without logic or regularity.

While this presumption captures some of the difficul-ties of finding a good design and, even more, of success-ful imitation, it is too strong. There is a logic underlyingthe idea of “fit.” Certain strategies and organizationaldesigns do fit one another and the environment, and thusproduce good performance, and others do not. Moreover,there are frequently recognizable, understandable, andpredictable relations among the environmental featuresand the choice variables of strategy and organizationthat determine which constellations of choices will dowell and which are less likely to do so. These relations

Key Concepts

33

arise for both technological and behavioral reasons.Recognizing these relations and understanding theirimplications can guide the design problem.

Further, when these relations are present there willoften be a quite restricted number of coherent patternsamong the choice variables. The design problem then isto identify and select among well fitting, often quite dis-tinct arrays of choices on the many dimensions. Thus,the problem becomes much more tractable when we havesome understanding of the factors that generate fit.

We have already seen one example of these ideas. TheHBC’s organizational policies fit with one another andwith the strategy and the environment (until the environ-ment changed with the entry of the NWC). So too did theNWC’s choices. They each had an internal logic to whichthey adhered in all crucial dimensions. Moreover, “mixand match” would not have worked. For example, perfor-mance pay would not have been particularly valuable atthe old HBC, given that the range of allowed behaviorswas tightly circumscribed. Thus, there was really a lim-ited set of alternatives that were apt to be of interest.

But why are there a limited number of coherent patterns?What defines them? Why does mix and match not work?

The idea that strategy and structure need to fit with oneanother and with the business environment is an old one, asis the recognition that there may be several distinct patternsamong these variables that are coherent, but not necessarilyequally good. These ideas have, however, rarely been for-malized. Recent developments in economics allow doing soin a simple, intuitive, and powerful way. The key ideas arecomplementarity among choice variables, non-convexity in theset of available choices, and non-concavity in the relationship

Key Concepts

34

between choice and performance. Complementarity givesrise to clear patterns of coherence in design. Non-convexityand non-concavity mean that there can multiple coherentpatterns that are quite distinct. Together these ideas givegreat insight into the problems of designing and changingan organization. In this context we will especially exploresome of the difficulties of organizational change. This leadsto considering another feature, the extent of tight couplingin the design, which reflects the extent to which theorganization is finely tuned to maximize performanceagainst a particular strategy and environment or, instead,is designed to work reasonably well in the face of change.

Complementarity

Complementarity involves the interactions amongchanges in different variables in affecting performance.Consider any pair of variables that the designer mightdetermine or influence in attempting to realize the firm’sgoals. Prices, service levels, frequency of productredesign, the debt–equity ratio, intensity of performancepay, the allocation of decision authority to subordinates,and aspects of the culture are all possible examples ofsuch choice variables. Then the two choice variables arecomplements when doing (more of) one of them increasesthe returns to doing (more of) the other. In more mathemat-ical language, the incremental or marginal return to onechoice variable increases in the level of any complemen-tary choice variable.1 Thus, if one of a pair of comple-ments is instituted or increased, it will be more attractivethan before to introduce or increase the other.

For example, price and product quality are comple-ments if higher quality makes demand less sensitive toprice increases (less elastic). Then an improvement in qual-ity renders a higher price more attractive because a priceincrease leads to a smaller fall in the quantity sold than itotherwise would.

In contrast, activities are substitutes if doing (more of )one reduces the attractiveness of doing (more of ) theother. For example, direct monitoring of employees’behavior and the use of performance-based incentive paymay be substitutes. If introducing performance pay givesstronger incentives for good behavior, because the resultsof this behavior are rewarded, then value of monitoringto enforce the desired behavior directly is probably lowerat the margin, and the level should be reduced.

Another example of substitution concerns make-to-stockversus make-to-order. Producing output in response to spe-cific customer orders is a substitute for making to stock,where output is produced in advance of receiving ordersand held in inventory until the demand materializes(Milgrom and Roberts 1988a). The claim here is not justthe obvious one that these two approaches seem to bealternatives. Rather, if we look at the fraction of output pro-duced under each regime, then the higher is the fractionmade to order, the more attractive it is to increase this frac-tion even more (and, correspondingly, to lower the fractionmade to stock). The reason is that producing to stock is sub-ject to economies of scale with regard to the level of inven-tory needed to maintain reliability in meeting demand as itemerges. Consequently, if it is worthwhile producing a littlebit to stock, it is even more worthwhile to do the next incre-ment (and correspondingly cut the share made to order).

Key Concepts

35

These ideas of complements and substitutes can beextended to relations among aspects of the environmentand the designer’s choice variables. For example, a choicevariable is complementary with an element of the envi-ronment if an increase in the level of the environmentalvariable increases the returns to introducing or increasingthe choice variable. So, if income tax rates are lowered, itmay be more attractive to increase the use of explicitperformance pay, since the cost to the firm of providing agiven intensity of incentives is reduced. Performance payand the negative of the tax rate are complements. Theideas can also be extended to groups of choice variables:Several choice variables are complementary if each ofthem is a complement for each of the others.

The substitution and complementarity relations amongchoice variables give structure to the problems of organiza-tional design. In particular, complementarity results invery clear patterns, with all the complementary choice vari-ables tending to be done together and at comparable levels.

This may be intuitively clear, but to see the logic moreclearly, consider another example of complements: theflexibility of a firm’s manufacturing system and the varietyof its product offerings. We might measure flexibility bythe speed with which the firm can change over from pro-ducing one product to another, or by the (inverse of the)cost of changeovers. Variety could be represented either bythe breadth of the product line at a point in time or by thefrequency of product changes. In any case, flexibility andbreadth should be complements in normal circumstances.Broadening the product line presumably increases thetotal demand facing the firm but lowers the potentialsales of each individual product as customers sort

Key Concepts

36

themselves among the wider set of alternatives. Unlessaverage inventory levels are to increase dramatically, thismeans shorter production runs and more frequentchangeovers. This, in turn, increases the value of beingable to do changeovers more quickly and cheaply—ofmore flexibility. (In general, doing more of any activityand lowering the activity’s marginal cost are always com-plements.) So having more variety increases the returns toincreased flexibility, and the two are complements.Conversely, a more flexible production system lowers thecost of realizing the demand advantages of having abroader product line, so the relationship holds in thatdirection as well. This symmetry is no accident—it isalways true that if the returns to increasing one variableare nondecreasing in the level of a second variable, thenthe returns to increasing the second are also nondecreas-ing in the level of the first.

Complementarity gives rise to systems effects, with thewhole being more than the sum of the parts (in a precisesense). Consider increasing each of a collection of choicevariables. Suppose we look at the effect on performance ofincreasing each one in isolation, without raising the others,and then sum these estimates. Then complementarityamong the activities means that the total effect on perfor-mance of increasing all the variables together exceeds thesum of these individual impacts. This is because comple-mentarity means precisely that, once we have raised thelevel of one of the activities, the impact of raising any ofthe others is now greater than it would have been whenthe first variable was at a lower level.

Indeed, when the variables are complements, it is quitepossible that changing any one of them alone would

Key Concepts

37

Key Concepts

38

worsen performance, yet changing all together wouldincrease it substantially. In this regard, Eric Brynjolfssonand his colleagues2 have studied the impact on productivityof the huge investments of the last decades in informationtechnology. They find that the costly investments hadlittle impact on productivity on their own, which matchesNobel Laureate Robert Solow’s quip that one sawcomputers everywhere except in the productivity statis-tics. However, when the investments are matched withcomplementary changes in the organization design, thereare very significant productivity effects. Neither theinvestments nor the organizational changes might beworthwhile on their own, but together they have a hugepositive effect on performance.

Coherence among a set of complementary choice vari-ables tends to result in all of them being set at a high levelor all at a low level. Consider the flexibility and varietyexample. When we allow for choices to be either highor low on each dimension, we might expect that therewould typically be two coherent choice patterns in thevariety/flexibility problem: Lots of variety and lots offlexibility, or little of either. This is because it will beworthwhile to bear the costs of flexibility only if thedesired variety is high, and a high level of variety will beworthwhile only if the production system is quite flexible.

The automobile industry provides an example of eachpattern (see Figure 2). Famously, in the first decades of thetwentieth century, Henry Ford would sell you any car youwanted, as long as it was a black Model T. Thus, Ford’sproduct line was very narrow, and it did not changeoften—Ford kept the Model T in production for decades.On the flexibility side, Ford’s plant was completely

specialized to the Model T. It was so inflexible that it hadto be gutted when the switch was finally made to theModel A. These features of organizational design andstrategic choice obviously fit with one another.

On the other hand, Toyota in the last decades of thetwentieth century had extremely flexible factories and abroad product line. For example, in the early 1990s Toyota’sKamigo engine plant on any given day produced over 350different engine/transmission/fuel-system combinations(including both single and dual camshaft engines) on a sin-gle line, in batches of one—each successive item coming offthe line was different from the one before. The product linewas broad, and the factory was very flexible.

Each of these patterns has an internal logic. Further,each was, in its environment, arguably optimal. Ford’sstrategy allowed it to build and dominate an industry: Atone point more than half the cars in existence were FordModel Ts. In the 1990s, when flexible automation wasmuch more readily available than it had been in the earlyyears of the century and when customer tastes hadbecome more diverse, Toyota was viewed as the leadingfirm in the global auto industry and as perhaps the bestmanufacturing firm in the world.

While there may be multiple coherent patterns forcomplementary organizational features, what typicallydoes not work is “mix and match” among elements of dif-ferent patterns. The high variety/low flexibility combi-nation is probably not even worth considering in mostmanufacturing contexts. Either production runs wouldbe kept short, implying immense costs from the frequentchangeovers, or long production runs would be employedto avoid costly changeovers, implying that the firm would

Key Concepts

39

Key Concepts

40

face financing gigantic inventories. The other off-diagonalpoint is less problematic in some ways, but can be dis-astrously expensive. It is, in essence, where GeneralMotors found itself in the 1980s. In that decade GMspent more than the combined market values of Toyotaand Nissan on flexible automation and other, related cap-ital expenditures that increased potential flexibility. It did not, however, adequately speed up its product devel-opment processes, adjust its product mix and productionscheduling, reform its human resource practices, and takea variety of other actions that are complementary withincreased production flexibility. Indeed, its assembly linesstill often produced only a single model of car, even whenthe machinery would have permitted producing severalmodels on a single line. GM at the end of that decade seta new record for the amount of money lost by any corpo-ration in one year, and then it broke the record the nextyear. While other factors contributed to this disaster,GM’s long, painful decline through the 1980s was insignificant part due to this mismatch in its organizationaldesign choices.

Low High

Fle

xibi

lity

Variety

High

LowFord

Toyota

?

GM

Figure 2. Complementarity can lead to discrete coherentpatterns

In fact, GM was far from alone in making thesemistakes. For example, Jaikumar’s work (1986) on theadoption and use of the innovation of computer numeri-cally controlled (CNC) machine tools in the UnitedStates and Japan gives another example of failure to adaptfully. He found that Japanese firms had rapidly seen thecomplementarity between the flexibility afforded by theCNC tools, shorter production runs, and greater variety.They were producing an increased variety of products invery small batches. In the United States, however, manyfirms initially used the highly flexible manufacturingequipment just as they had used the older, inflexiblemachines, to produce huge numbers of single items.

Another, richer example of complementarity amongchoices involves the unique set of practices that mark theLincoln Electric Company (Berg and Fast 1975, see alsoMilgrom and Roberts 1995). Lincoln Electric’s primaryproduct is arc-welding equipment and the consumables,such as flux, used in welding. Lincoln long dominated theindustry in the United States, leading major firms such asGeneral Electric and Westinghouse to exit rather thanattempt further to compete with it. Moreover, Lincolnenjoyed an unbeaten record of performance—it was prof-itable every quarter, year in and year out, for almost a cen-tury, and its U.S. operations have never lost money or hada layoff. The basis for Lincoln’s success was its executionon a strategy of pursuing ever higher productivity andlower costs and then passing on some of these benefits tocustomers in lower prices. It was able to improve produc-tivity and costs so consistently because of a set of organ-izational policies that supported these aims. Together,there was a high complementarity between its strategy

Key Concepts

41

and organization and among the elements of its organiza-tion design.

The centerpiece of Lincoln’s organizational design isvery extensive use of piece-rate compensation. Wheneverpossible, workers are paid piece rates, that is, a fixedamount for each unit on which the individual workercompletes the assigned tasks. Indeed, at one point typistswere even paid by the keystroke (until one was discoveredsitting her desk at the lunch break, repeatedly hitting thesame key while she ate) and overhead crane operatorswere paid by the number of objects moved (until safetyconcerns arose about the speed and height at which heavyitems were being moved around the plant).

Piece rates give very strong, straightforward incentives,not just to work hard but also to seek ways to increaseone’s output further. They are easy to understand andadminister. At one time they were a very common methodfor paying industrial workers. Yet they are now a compar-ative rarity. There are several reasons for this. First, if thepace of work cannot vary between individuals, then thepossibility for responding to the piece-rate incentives islimited and they are not of much use. Thus, individual piecerates have little value in such contexts as assembly lineoperations or team-organized work. The second difficultyis that by giving strong incentives for quantity, a piece-ratesystem discourages spending time and effort on othervaluable activities. Most directly, if skimping on qualityallows the worker to increase the number of pieces made,there is a perverse quality incentive from piece rates.Moreover, other desirable activities that cannot be paid atpiece rate—such as helping other workers or acceptingtemporary reassignments to deal with emergencies—are

Key Concepts

42

discouraged. The third difficulty is that workers oftendistrust managers and fear that the piece rates will be low-ered if they respond fully to the incentives that the piecerates appear to offer and they thus reveal just how pro-ductive they can be. Consequently the piece rates are notas effective motivators as they would otherwise be. (Thismay in part explain why unions have usually opposedpiece rates: They fear that some workers will respond andthen all will be subjected to higher standards.) Finally,there is the problem of selling all that might be made andof maintaining the system when sales slump.

Lincoln responds to each of these difficulties with poli-cies and processes that make the piece rates more effec-tive. The production system is designed to allow work tobe individually paced, so that workers will have the free-dom to increase their output rates in response to theincentives. To facilitate this, significant amounts of work-in-process inventory are tolerated. To overcome the“multi-tasking” problems with quality and cooperation,Lincoln uses an individual bonus scheme. The amount ofthe individual bonus is determined by the employee’ssupervisor and is based on the quality of the employee’soutput and such factors as perceived cooperativeness.These bonuses normally double the employee’s baseearnings from the piece rate. In addition, each employee’sname is stencilled on each welding machine on which heor she works, so that responsibility for quality problemscan be assigned. If a piece is found to be defective oninspection, the responsible worker must repair it on his orher own time. If a machine fails in the field as a result ofa worker’s skimping on quality, the worker’s bonus isdocked by as much as 10 percent.

Key Concepts

43

Lincoln has also adopted a whole range of organizationalfeatures that help overcome the trust problems and therebymake the piece rates credible. First, the company promisesthat the rates will not be adjusted unless there is a changein the product or work methods, and it allows an employeeto challenge any new rates and have them recomputed.3

What makes this promise credible, however? First, the firmis essentially employee-owned. For a long time, theemployees and managers and the founding Lincoln familyheld most of the stock, and when Lincoln issued newequity to the public in the mid 1990s, it was offered with-out voting rights. This ownership arrangement reduces thedanger than investors who do not understand the value ofthe commitment or are too impatient will force reductionsin the rates. Second, early in its history the companyadopted a set of measures that encourages two-way com-munication between workers and management. Suchschemes are not uncommon today, but they were an inno-vative rarity when Lincoln adopted them. Third, Lincolnwas run by the founder and his brother until 1965 andthen, for the next three decades, by career Lincoln employ-ees. They were personally committed to the system, andthey well understood its logic and the need to honor theworkers’ trust. Fourth, Lincoln has a number of policiesthat are symbolic of the relative positions of managementand workers, including no assigned parking spaces forexecutives, low executive compensation levels, no separateexecutive dining facilities, and Spartan managerial offices.These factors increase employees’ trust of the managementand reduced the danger of an “us versus them” mentality.

Finally, to deal with the problems of matching outputto demand, Lincoln normally rations the time that

Key Concepts

44

workers can be at their posts while demanding mandatoryovertime when it is needed.

Each of these features of Lincoln’s design may or maynot be attractive on their own, but together they are power-ful because they complement one another. Work-in-processinventory is often viewed as anathema in modern, “lean”production systems, but Lincoln’s tolerating it permits theflexibility in the pacing of individuals’ work that must bepresent to allow workers to respond to the piece rates.Absent offsetting incentives for quality and cooperation,piece rates can be disastrous, and so Lincoln’s piece rates aremade more effective by the complementary bonus scheme.The effectiveness of the piece rates is also increased by thehigh levels of trust that are supported by the open commu-nication policies; the ownership structure; the long-term,internally recruited leadership of the firm; and the symbolicacts and policies. The relative rarity of some of these fea-tures suggests that they are not valuable in more standardmodels of organization. They are, however, importantelements of the model that underlies Lincoln’s success.Finally, Lincoln attracts and retains employees who like theLincoln model, and this makes it more effective. Judgingfrom interviews with Lincoln’s employees, they are orientedtowards material success and are willing to work hard toachieve it, they are attracted by the individual responsi-bility and autonomy that Lincoln offers, and they do notwant a union.

Lincoln’s organization permits it to achieve remarkablyhigh levels of productivity, which is key to realizing itslow-cost strategy. But Lincoln actually aims not just forlow costs, but also for constantly reducing costs. A majorpotential barrier to doing this would be the workers’

Key Concepts

45

concern that increasing productivity might endangeremployment, because the same output could be producedby fewer people. This helps explain another of Lincoln’svery unusual policies, a promise to avoid layoffs. It hasstuck by this promise even in severe recessions, when itput production workers to painting the factory ratherthan lay them off. This policy also probably contributesto the trust that supports the credibility of the piece rates.So another policy that might be of questionable attrac-tiveness in isolation becomes very valuable in the contextof what Lincoln seeks to do and how it seeks to do it.

Thus, the features at Lincoln together yield muchmore than would be estimated by looking at the impact ofdoing any one of them in isolation, exactly because theyare complementary with one another.

As this example suggests, the range of variables overwhich complementarity can spread can be very broad andthe corresponding patterns very rich. Other examplescan be found in the literature, especially in the context ofhuman resource management (HRM). Pfeffer (1996) andBaron and Kreps (1999) have argued for the existence ofcomplementarities among rich sets of HRM practices.Baron, Burton, and Hannan (1996) have found evidencefor these complementarities in the practices of SiliconValley start-ups, identifying limited numbers of patternsthat are actually adopted out of the thousands that areconceivable. Ichniowski, Shaw, and Prennushi (1997)have documented patterns in the HRM systems used inspecialty steel finishing lines that are best explained interms of widespread complementarities among manu-facturing and HRM practices. Bresnahan, Brynjolfsson,and Hitt (2002) have found complementarities among

Key Concepts

46

Key Concepts

47

investments in information technology, workplaceorganization, product innovation, and a number of HRMpractices in a large sample of U.S. industrial firms.

Perhaps the broadest set of complementarities studiedso far involves traditional mass production versus modern,lean manufacturing (Milgrom and Roberts 1990b, 1995).4

Mass production and lean manufacturing represent twocoherent patterns of choices over a very large set of policyvariables, where a move of any one element from the massproduction model practice to the lean model is comple-mentary with the corresponding move on each of theother variables. (See Tables 1 and 2.)

While examining all the interactions among such a richset of variables would be too much of a digression, we cansketch some of them. The connections between the flexi-bility of capital equipment, the length of production runs,the levels of inventory, the breadth of the product line, andthe frequency of product changes have already been dis-cussed. The approach to marketing and to communicatingwith customers is then driven largely by need for them tofit with the product strategy. The frequency of changes inproducts drives the value of speed in the product develop-ment process. Drawn-out, sequential approaches to prod-uct development work in the Mass Production model butare infeasible in the Modern model, which requires theuse of cross-functional teams to get new productsdesigned and manufactured speedily. Frequent changes inproducts and frequent process innovations favor having ahighly skilled workforce that can both handle the com-plexity and solve problems as they emerge (rather thanwaiting for the managers and engineers to do this), so abil-ity and training are complementary with innovation.

Key Concepts

48

To take full advantage of the workforce’s abilities, it is thenmore worthwhile to elicit their involvement in developingprocess improvements, so empowering the workers andlooking for continuous improvement are complementarywith their being more skilled. With skilled workers whoare particularly knowledgeable about the firm’s activitiesand needs, it is desirable to establish long-term employ-ment relations in order to help retain this valuable humancapital. Meanwhile, the flexibility of the manufacturingequipment reduces the dangers of lock-in to particular

Table 1. Characteristic features of mass production

Logic: The transfer line, interchangeable parts, andeconomies of scaleSpecialized machineryLong production runsInfrequent product changesNarrow product linesMass marketingLow worker skill requirementsSpecialized skill jobsCentral expertise and coordinationHierarchic planning and controlVertical internal communicationSequential product developmentStatic optimizationAccent on volumeHigh inventoriesSupply managementMake to stock, limited communication with customersMarket dealings with employees and suppliersVertical integration

Key Concepts

49

suppliers and customers and so favors external sourcingrelative to internal supply. This is further supported by theadoption of long-term relationships with suppliers.

Modern, lean manufacturing is supplanting massproduction in many industries and having strongly positiveeffects on performance. This is an example of one coherentpattern of strategic and organizational choices becomingmore effective than another that had previously seemedbest. Such changes occur in response to changes in the envi-ronment. Complementarity is also valuable in understand-ing when different patterns will be more or less effective.

Table 2. Characteristic features of modern manufacturing

Logic: Flexibility, speed, economies of scope, and corecompetenciesFlexible machines, low set-up costsShort production runsFrequent product improvementsBroad product linesTargeted marketsHighly skilled, cross-trained workersWorker initiativeLocal information and self-regulationHorizontal communicationCross-functional development teamsContinuous improvementAccent on cost and qualityLow inventoriesDemand managementMake to order, extensive communications with customersLong-term, trust-based relationshipsReliance on outside suppliers

When choice variables are complementary, anyenvironmental change that increases the attractiveness ofraising one of the variables tends to result in all of thembeing increased. This gives rise to systematic, predictablepatterns in how the choices move in response to environ-mental changes. To see this, suppose, for example, that aset of initial choices among a group of complementaryproductive inputs actually maximizes performance. Thismeans that, in particular, no decrease or increase in anyof the input levels from this point is worthwhile.Suppose then that the cost of one of the inputs falls byenough that it is now worthwhile to raise this variable.But doing so raises the attractiveness of an increase ineach of the complementary choice variables, and so weexpect that all the other input usage levels will increase inresponse to the increase in the first one. This, in turn,raises the returns to increasing the level of the first inputfurther, and doing so again increases the returns to rais-ing the others. The final result is that all the choicevariables have increased in response to an environmentalshift that initially favored increasing only one of them.

Applying this logic, the move from mass production tolean manufacturing that has been occurring in the lastdecades could be a response to a number of environmen-tal changes. Certainly the cost of flexible manufacturingequipment—computer-aided design and manufacturingequipment, numerically controlled machinery, and indus-trial robots—has fallen: Indeed, these did not exist untilrecently. The first effect of this is to encourage the use ofmore flexible manufacturing systems, and the adoptionof these then favors the other elements of the overall shift.Also, the cost of communicating with customers and

Key Concepts

50

Key Concepts

51

suppliers has fallen with improvements in communicationand information technology, which favors a make-to-orderapproach and, indirectly, all the features that are comple-mentary with it. It may also be possible that, with risingincomes, consumer tastes have shifted towards greatervariety, and this too would favor the move to the newmodel. Again, the increases in the level of formal educa-tion in the workforce that have occurred over the decadeswould favor a shift to making more use of workers’ brains(and not just their brawn), and this too is complementarywith the other elements of the lean model. Finally, asa firm’s suppliers and competitors move to the new model,its incentives to do the same are increased.

This raises the issue of moving between patterns ofstrategic and organizational design change. The mathe-matics of complementarity indicates that, in a system ofcomplements, any change in the environment that favorsincreasing a particular variable then also leads to increasesin all the other variables. Thus, when the choices are com-plementary, the direction of desired change is unambigu-ous. Yet, organizational change seems, in fact, to be verydifficult. The reasons why this is so are many, but oneelement of the answer lies in understanding more deeplythe nature of the relationship linking the environment,the strategy, and the organizational design to perfor-mance. Here is where the second idea, failures of convex-ity and concavity, comes in.

Non-convexity and Non-concavity

Traditional models in economics and the social and man-agerial sciences have usually made particular mathematical

assumptions—formally, convexity of the choice set andconcavity of the objective—that were borrowed from thephysical sciences. They were used because they facilitatedanalysis of the models by the methods of differential cal-culus. Such models often underlie the intuition that bothscholars and practicing managers use in trying to under-stand the world. Yet, these mathematical assumptions areoften quite inappropriate when dealing with problems ofstrategic and organizational choice, and the intuition theyyield often is quite wrong. Replacing them opens upimportant new insights.

The first assumption, convexity of the set of alterna-tives, is that if two options are available, then any inter-mediate choice is also possible. In particular, thisassumption implies that choices are infinitely divisible.The second assumption, concavity of the objective func-tion, deals with nature of the relationship between choiceand performance for a given environment. In the casewhere choice is represented by a single variable, concavitymeans that the impact on performance of successiveincrements to the choice is decreasing, perhaps ultimatelybecoming negative. More generally, it requires that if twodistinct choices lead to the same performance, then anychoice intermediate between the two would lead to ahigher level of performance. These properties (plus someboundary conditions that were also typically assumed)imply that there is a unique performance-maximizingchoice for any environment.

This sort of situation is illustrated in Figure 3, wherethe relation is graphed between choice (assumed to beone-dimensional to allow graphical representation) andperformance for a given environment. As the choice

Key Concepts

52

Key Concepts

53

increases, the realized performance increases, reaches amaximum at choice marked X, and then declines.

The first implication of this sort of modeling hasalready been noted: There is a single best way to do things(denoted X in the figure). Further, if the current choice isnot at the best point, then small changes in its directionwill improve performance. For example, at the point Y inFigure 3, any move towards X improves performance.More generally, if there is some other point that offersbetter performance, then a small adjustment in the choicein the direction of the new point, will be performanceenhancing. Thus, the task of finding the best choice is rel-atively simple, because local experiments that improveperformance will eventually lead to the optimum. Thisremains true even when there are multiple dimensions tothe choice. When choice is multidimensional, concavitymeans that the optimum can be found by very uncoordi-nated, local, decentralized experimentation and search.Change any element of the design, on its own, by a smallamount in a direction that increases performance, andcontinue to make such adjustments as long as any arepossible. Then when the process stops, the performance-maximizing design has been found. So there is no

Perf

orm

ance

ChoiceX Y

Figure 3. The classic model of choice and performance

coordination problem in decentralized search forimprovements in performance.

A second, subtler implication concerns the task ofmaintaining the optimal choice as the environmentchanges. Suppose changes in the environment reshapethe connection between choice and performance, so thatthe optimal choice changes, but that the general concaveshape of the relationship is maintained. Then it seems arelatively simple matter to track the changes and keepnear the moving target. A little local experimentation,kaizen-like, will reveal whether an improvement is possi-ble and in what direction to change. Start moving thechoice in that direction, and when no further improve-ment is possible, the new best point has been found.

Suppose, for example, the initial relationship betweenchoice and performance is shown by the solid curve inFigure 4, and the firm manages to put itself at X, the opti-mal strategy and organization. Now suppose the environ-ment changes, so that the dotted line reflects theconnection of choice to performance. Performance at Xhas not changed in this example, although in general itmight have. The new optimum is at X�, involving a higherlevel of the choice. Experimenting by changing the choicejust a little at X signals that the new optimum lies athigher levels, because increasing the choice increases per-formance, and decreasing it hurts performance.

The problem with such models and the intuition theygenerate is that the maintained assumptions underlyingthem are quite implausible in thinking about strategic andorganizational choice. Indivisibilities abound (the firmcannot have a fractional number of plants; it either entersa market or it does not). These are inconsistent with

Key Concepts

54

Key Concepts

55

convexity. Further and more crucially, increasing returnsto scale, learning effects, and indivisibilities are all incon-sistent with concavity of the objective. For example, withincreasing returns, zero profits would result both fromnot operating at all and from operating at some positive,break-even level of output. Yet, operating at an interme-diate level would produce losses, because the costsper unit are higher there. This directly counters thedefinition of concavity, so performance is not a concavefunction of choice. In fact, if there are multiple elementsto the choice and these show strong, pervasive comple-mentarities, then mathematically it is impossible to havethe conditions that underlie these models.

A number of important insights into managementproblems follow from recognizing these simple facts andembracing the possibility, and even the likelihood, thatconvexity and concavity will not hold.

The first is a basis for understanding why some firmsseem to be constantly changing their organizations, goingfrom centralized decision-making to decentralized andthen back again, apparently aimlessly. This has puzzlednumerous observers of organizations and has been raised

XChoice

Perf

orm

ance

X�

Figure 4. Shifting the relationship between choice andperformance

as an objection to the very idea that organizational designis actually done in any rational way. In fact, this can be avery effective response to indivisibilities.

The key lies in recognizing that one who has theauthority over a decision is inherently indivisible; eitherthe authority rests with headquarters or it is delegated toan operating unit. Now suppose that the organizationaldesigner cannot directly control some of the less formalelements of the design, like the operation of personal net-works or some of the elements of corporate culture.Instead, these evolve in ways determined by the formalaspects of the organization, including the allocation ofdecision authority. In particular, suppose that if decision-making is centralized certain norms weaken, while thereverse is true under decentralization. These normsmight govern behavior like risk-taking or the allocation ofeffort among tasks. Finally, suppose that the norms arewhat really matters for performance and that the best per-formance comes when the norms are at some middlinglevel of strength, between the limits to which they wouldtend under permanent centralization or permanentdecentralization. Then if organizational change is costly,the optimal solution is to alternate intermittently betweencentralized and decentralized decision-making. When thenorms drift too far in the direction induced by the currentallocation of decision-making authority, performancestarts to suffer. The solution is to switch to the other allo-cation, thereby reversing both the drift in the strength ofthe norm and the declining performance. The analogy iswith a furnace being only on or off, and the temperaturein the house drifting in the direction determined by thestate of the furnace. When the house gets too hot, the

Key Concepts

56

thermostat shuts off the furnace, and when the temperaturefalls far enough, it turns it back on.5

A second, more fundamental, major insight is thatthere may be multiple choices (or patterns of choice,when there are many choice variables) that are coherent.Yet, among these multiple coherent patterns, some mayyield much better performance than others.

Here “coherence” has a double meaning. First, itrequires that no small adjustment in the set of choices canincrease performance—the choice is “locally” best. So if achoice is coherent, then kaizen-type seeking for a little bet-ter way to do things, even if coordinated across decision-makers, yields no improvement (unless the environmentchanges). Second, when choice is multidimensional, thenno change, however big, in just some proper subset of thechoice variables can improve performance. Thus, if theorganization is at a coherent point, even if it shows poorperformance, it is still possible that managers cannot find abetter solution unless every element of strategy and orga-nizational design is changed in a coordinated fashion.

The first sort of coherence—local optimality—is illus-trated in Figure 5. The horizontal axis shows the choice(assumed again to be one-dimensional to allow graphicalrepresentation), and the vertical the resulting perfor-mance. The point Z is not coherent—indeed, any smallchange from Z improves performance. There are two dis-tinct points, X and Y, from which no small change yieldsan improvement and so which meet this local optimalitycriterion for coherence. Yet, Y clearly has better perfor-mance. This could not happen if the relationship betweenchoice and performance were concave, because then therecould be only one local maximum.

Key Concepts

57

Key Concepts

58

To illustrate the second aspect of coherence, considera situation where organizational performance is the min-imum of the performance levels of the different units inthe firm. This is essentially the case when output is pro-duced by different people or units working in sequence,each doing an operation, as in an assembly line. (Note thatin such a situation, the actions are complementary: Oneperson working harder is more valuable if others dothe same.) Then any pattern where all the units are alloperating at the same level has the second element ofcoherence: No change, however large, in any subset of theactivity levels can increase performance. This is becausesome of the activities will not have been changed and so,even if we increase all the others, the minimum is stillthe same, and overall performance is unchanged. Forexample, if there are three organizational units, all oper-ating at level 2, then the organizational performance ismin{2, 2, 2}�2. Increasing the first two activities to level 3still leaves organizational performance at 2�min{3, 3, 2}.Actually increasing performance requires all the activ-ity levels to increase. (Note, however, that small coordi-nated changes can be performance enhancing in this

Perf

orm

ance

ChoiceY XZ

Figure 5. A non-concave performance relationship

Key Concepts

59

set-up, so we do not necessarily have the first element ofcoherence.)

More dramatic, and more important managerially, aresituations where there are multiple patterns that meetboth tests for coherence. To illustrate these we need toconsider several choice variables—at least two. Ratherthan drawing the relationship between the two dimen-sions of choice and performance in a three-dimensionalpicture, we look just at a contour map, as in Figure 6. Themap there shows a “mountain” with two “peaks,” one tothe northeast of the other, and a “high pass” across aridge between them. The curves connect choices withequal performance, and the arrows point “uphill.”

There are two points here that are local maxima, eachof which lies inside one of the smallest oval contour lines.One (marked A) has relatively low levels of the choicesand the other, B, has both choices at higher levels. Theyare the peaks. Each of A and B has the first element ofcoherence in that any small change, even in both choice

A

A B

B

Figure 6. Non-concavity allows multiple, distinct coherentpoints

variables, lowers performance. They each also have thesecond property, that no move parallel to an axis (i.e. nochange in just one of the choices, no matter how large)can improve performance. Yet, either peak might havea higher level of performance than the other. If the firmfinds itself at the wrong one, then any improvementrequires large, coordinated changes on all (i.e. both)dimensions of choice.

An important implication for management is that,when complementarities and non-concavities abound,decentralized local experimentation is not enough.Search and change must be coordinated. This does notmean they must be driven from above, in a command andcontrol way. But leaving individual managers in charge ofparticular elements of the organization to find improve-ments on their own can fail miserably, as can experimen-tation that is limited in scope. Both can fail to find thebetter solution and instead leave the firm stuck at an infe-rior coherent point. This means either that realizing thebest design must be centrally coordinated—there needsto be a designer—or else that the different parties makingthe choices need to communicate intensively with oneanother.

Why or how would a firm end up on the wrong peak?One answer is that it simply made the wrong choice. Adeeper answer is that it can get trapped in the inferiorposition by environmental change. For example, the envi-ronmental change brought on by the entry of the NWCsuddenly rendered the HBC’s model inappropriate, eventhough it had worked very well for more than a century.Strikingly, however, the change need not be radical ordiscontinuous. Continuous change in the environment

Key Concepts

60

can, in the presence of non-concave performancefunctions, result in discrete, substantial changes in theoptimum configuration of strategy and organization.

Consider Figure 7. The vertical and horizontal axesare as in Figures 3 through 5 and show choices and theresulting performance. (If possible, however, think ofchoice as being multidimensional, as in Figure 6.) On thethird axis is time. Over time, as technology, the behaviorof competitors, suppliers, and customers, or other factorsgradually change, the curve linking choice and perfor-mance changes too. The diagram plots the relationship atthree points in time. Initially, at time T1, the choice of Y1is best. X1 is a coherent choice, but worse than Y1. Astime and change gradually proceed, the two coherent pointsmove somewhat, from Y1 to Y2 and then Y3, and from X1

Key Concepts

61

Y2 X2

Performance

T1

T2

T3

TimeX3

Y1Choice

X1

Y3

Figure 7. A shifting, non-concave performance relationshipresults in discontinuous change in the optimum

to X2 and X3. A firm that started at Y1 could probablytrack the change in the locally optimal configuration andadjust its strategy and organization on an ongoing basis,moving to Y2 and then to Y3 without too much difficulty.But note that, over time, the attractiveness of the Y posi-tions is slipping relative to the X’s. It need not be that Y3has lower performance than Y1 offered, although thismight be the case. But if the technological, competitive,regulatory, or social changes favor the X configurationenough, then it will pass the Y’s in performance terms.

In the figure, at time T2, the X2 configuration is as goodas Y2. By T3, X3 is clearly better than Y3. A firm that wasfollowing its basic Y strategy, skillfully adjusting it asneeded at the margins, would then suddenly find that itwas poorly adapted to the environment: There is a muchdifferent way, and it is now a much better way. The firmis then faced with sticking with its old ways, which maynot ultimately be viable, or undertaking radical strategicand organizational change.

Such changes are not easy, which can be seen as one ofthe reasons for dysfunctional organizational inertia, that is,failure to adopt changes that would seem worthwhile. Tobring them about requires several elements of leadership.

First, there must be strategic recognition. The mostbasic problem is to recognize the need or opportunity forchange. The HBC’s leaders failed to appreciate the needfor change to meet (let alone deter) the entry of theNWC. Similarly, Jack Smith, the former CEO of GeneralMotors, has said that the automaker was “in denial”through the 1980s, failing, somewhat willfully perhaps, tosee that the Japanese competitors had fundamentallychanged the nature of the business, and thus to see the

Key Concepts

62

need for basic change in GM’s strategy and organization.Here, past success becomes a trap.

The second required element of leadership is vision, inorder to see the other, better pattern, at least in broad out-line. This first requires an understanding of what sort ofchange is required. When American auto company man-agers and engineers went to Japan in the late 1970s andearly 1980s, they saw lots of differences between practicesin the Japanese auto plants and theirs. There were no vis-ible work-in-process inventories; the workers could (anddid) stop the production line to deal with problems at anytime by pulling a cord; there seemed to be no rework areawhere mistakes in assembly could be corrected (therework area was typically a quarter of the floor space in atraditional U.S. auto plant); the day began with exercisesand a company song; workers wore uniforms; there wereteams for various purposes; and so on. Which of thesemattered? Why? It took a long time to figure this out.Lacking a good theory about which of these mattered andhow the pieces fit together, the American experts broughtback a few of the novel features, especially quality circles.These flopped resoundingly because they were not sup-ported by a commitment to quality, by workers’ beingempowered to experiment with and change work methods,and by guaranteed employment (so that workers did notneed to fear that improving productivity would hurtemployment), all of which are complements to gettingworkers involved in improving things.

What this points to is the need for theory. Of course, theproblem is even greater when organizational innovation isrequired, rather than just mimicry. Then the organizationmust engage in the search for a new way. The sort of global

Key Concepts

63

search that is involved in developing radically differentmodels has little in common with local search forimprovements within a given paradigm, and even firmsthat are very good at the latter may struggle with theformer. Still, theory can help by suggesting where to focusthe search.

In this regard, it is important to accentuate that the newway need not be understood in all its complexity from theoutset. Indeed, what we would expect is that, as opportuni-ties arise, companies adjust their policies and add activitiesthat are complementary with what they already have. Thus,they would evolve towards patterns of strategy and organi-zation that are marked by a high degree of fit amongcomplementary activities, even though initially no one fullycomprehended all the complexity of the design that finallyemerged. Lincoln Electric’s system, in fact, developed overmany years, with features that complemented what wasalready in place being added seriatim. The piece rates wereinstituted when the company started at the end of the nine-teenth century, the bonus system was implemented in 1934,and the no-layoff policy was not formalized until 1958.What was constant was the vision of James Lincoln, thecompany’s long-serving president, who believed strongly inproviding individual incentives to increase productivity.

Similarly, the Toyota Production System (TPS), whichis the basis for the modern, lean manufacturing model,developed incrementally (Ohno 1988). As postwar Japanemerged from defeat and destruction and Toyota sought toresume automobile manufacturing, the creators of TPS,Eiji Toyoda and Taiichi Ohno, saw that they could not com-pete on the basis of mass production, because the Japanesemarket was too small and the demands (for trucks, limos,

Key Concepts

64

taxis, etc.) already too fragmented. They then startedbuilding a new approach that might allow them to suc-ceed without the efficiencies of large-scale operations.The basic theme was to save costs through the ruthlesselimination of muda—waste—and the crucial insight wasto view inventory as muda. Efforts to reduce inventory ledto the just-in-time approach to parts supply. This bothraised the importance of flexibility and revealed the pos-sibilities for improving quality when problems could becaught immediately and corrected. Doing so, however,required empowering workers to act on local information.Eventually, the whole modern model was developed.

The third element of leadership needed in bringingabout change is the ability to communicate, to explain thenew way, what its general features are, and how to getthere. The leader cannot in fact design the new system inall its detail and implement it single-handedly. Peoplethroughout the organization must be involved. So theymust understand what needs to be accomplished. Withcommunication there must also be persuasion, to convincepeople of the need for change and of the gains that itwill yield.

Finally, there must be courage to try for the distant anddifficult and not to turn back when the transition is noteasy and performance actually suffers. Such performancedeclines are almost inevitable during a major changeoperation. First, unless the firm can nimbly and accu-rately leap from peak to peak, the path from one to theother must lead downward in terms of the maximumattainable performance. Refer again to Figure 5: If theorganization cannot instantly change from X to Y, butmust instead track over the points between, performance

Key Concepts

65

goes down. Such tracking would reflect an inability tochange people, processes, and culture instantaneously.Even if the firm can execute giant leaps (which meanschanging everything overnight, including values, norms,and reputations!), unless it is extremely accurate in itsjumps, it will likely come down some distance from theother peak. Then performance will not be all that wasexpected.

Moreover, there is reason to expect that a firm in themidst of a major change will not realize even the maxi-mum performance possible from its imperfectly alignedorganization. Thus, it will be below the curve, rather thanon it. First, the organization’s resources and attention arediverted from getting work done to figuring out what todo and how to do it. Second, even once “what to do” isdetermined, it is not likely to be done with full efficiency.Changing established patterns of behavior means learninghow to operate in new ways and how to communicate withnew people about different things. This is likely to be aslow, costly process. Third, resistance to changes thatendanger the positions, power, perquisites, and pay ofthose in the organization may further degrade perfor-mance. Change is almost always a threat to at least someof the people in the organization. It threatens to break theimplicit contracts that have guided past behavior, remov-ing the promised rewards; it upsets the established alloca-tion of power; it may destroy the value of establishedskills and positions; and it may put jobs and careers atrisk. In these circumstances, people may actively resistthe changes. At a minimum, they are likely to eat up a lotof their time and others’, worrying about what will happen.This makes undertaking change costly.

Key Concepts

66

Further, the prospect of change creates opportunitiesto influence the new distribution of resources, power, andrewards, and members of the firm will have every incen-tive to attempt to ensure that they and their allies do well.They will expend time and resources campaigning tohave the change structured to their advantage, and theywill misrepresent information to twist decisions in theirfavor. So simply putting change on the agenda createsinfluence costs that would otherwise not arise. This maymean that changes that would, organizational politicsaside, be worthwhile are actually delayed or avoided.

This logic suggests one reason that change is easier toaccomplish in a crisis situation. The gains to resistingchange or attempting to influence its shape and directionare dependent on the continued survival of the firm.Then, worsening prospects and the threat of collapse canreduce the influence costs occasioned by strategic andorganizational change. This means that changes thatwould not be worthwhile to attempt in good times willactually be implemented when survival is in danger(Schaefer 1998).

Tight and Loose Coupling

The difficulties of organizational change imply that thereis value in an organizational design that will perform rea-sonably well in a variety of environments, even if it is notperfectly adapted to the current context. This factorintroduces a trade-off in the design problem. In a staticcontext with a given environment, solving the designproblem largely involves recognizing complementarities

Key Concepts

67

among different features and adjusting the differentaspects of the design to take advantage of these. When theinherent uncertainty about the environment is recog-nized, the designer must also decide on how tightly tolink the different aspects of the design. A system that is“tightly coupled,” so that changing any aspect of thedesign or the environment will compromise performanceseverely unless numerous other aspects are also adjusted,may work very well if all goes according to plan. A tightfit among strategic and structural choices and with theenvironment can yield great performance. The dangercomes when an unintended or unforeseen change knockseverything out of alignment by forcing a change in somechoice variables or making the previous choices no longerfit with the environment. Performance then may sufferdramatically. Further, finding a new pattern may be diffi-cult, and implementing it even more so.

Thus, a less fully optimized, more “loosely coupled”design may offer flexibility and be favored when changesin the environment or autonomous change in the organi-zation are likely. Then adjustments can be made whereneeded and performance maintained, without incurringthe costs of massive restructuring of the whole system.

For example, a standard assembly line is the ultimateexample of a tightly coupled system: No one on the linecan reduce the pace of his work without affecting every-one else. Lincoln Electric’s design is more loosely cou-pled by virtue of the work-in-process inventories that itmaintains between different steps in the productionprocess. Holding these inventories is very costly, and theirlevels could be lowered significantly if everyone alwaysworked at a constant pace. However, they are useful

Key Concepts

68

because they permit individual variations over time in thepacing of work. The research program of the typical aca-demic department is almost entirely decoupled—whatone professor does, or does not do, can change with littleor no impact on the other faculty members’ activities. Ofcourse, there is little direct gain on the research side fromhaving these faculty grouped in the same department.

These ideas offer one explanation of the difficultiesthat Japan has been experiencing since the early 1990s.The system that marked the Japanese economy and majorJapanese corporations through the 1980s was character-ized by a dense, rich web of complementarities thatextended through demographic and cultural variables tonational policies and on to corporate design and manage-ment in an economy-wide example of tight coupling.6

The key element was a growth orientation in the wayfirms were run that fed off and supported the highgrowth and savings rates in the Japanese economy. Thisfocus on the long-run survival and growth of the firm inturn fit with the permanent employment policies, theheavy investments in human capital, and the reliance onlong-term supplier relations. The accentuation of growthover profitability was supported by the governance sys-tem, where shareholders had little power and boards ofdirectors were completely dominated by inside, executivedirectors; the financing by affiliated banks that wouldsupport the long-term orientation; and the restrictions onJapanese individuals’ investing their savings outside thecountry. Social attitudes legitimized the system, accentu-ating hard work, loyalty to the employer, the developmentof the nation from its postwar destitution, and savingsover consumption. Every element of the model was finely

Key Concepts

69

adjusted to every other, and together they gave an exquis-ite fit with the prevailing environment and produced out-standing economic performance for several decades.

Then a series of shocks hit. Japan reached the techno-logical frontier and could no longer rely on net imports oftechnology as a source of growth. Population growthslowed and the population started to age rapidly.Successful corporations accessed the Eurobond marketsfor financing, depriving the main banks of a key part oftheir traditional business, driving them into real estateloans. These helped create the bubble in asset pricesin the late 1980s and then went sour when the bubbleburst. Social attitudes changed in ways that reduced thewillingness to work for the good of the corporation andcountry. Foreign political pressures concerning tradeincreased, meaning that growth could no longer be so easilyachieved by exporting more. The system had presumedthe way things had been on all these dimensions, and nowthey had changed. The other elements—growth-orientedstrategies, permanent employment, firms being runessentially in the interest of their permanent employees,main-bank monitoring in place of stock market and out-side director oversight of management, and so on—nolonger fit, and performance has suffered now for a decade.But it is no easy task to change all the elements of thepattern to find a new one that works, especially in ademocracy, where no one is empowered to play the role ofdesigner. Indeed, it has taken a long time for leaders thereto begin to realize that the problems were not simplemacroeconomic ones, but fundamental structural ones.Japan is still struggling to find a new way.

Key Concepts

70

Firms, in fact, can vary the tightness of coupling ofdifferent elements of their organizational designs. Someelements, like financial controls or IT platforms, could beabsolutely standardized throughout the organization,with no variation permitted. Meanwhile, marketing orhuman resource practices might be allowed to vary acrossdifferent geographies to respond to local variations intastes, market conditions, and regulations. Such varie-gated coupling is, in fact, probably best. The issue iswhich parts to couple tightly, which to allow to be com-pletely decoupled, and how loose the loose couplingshould be to permit adaptation to change.

The tightness of coupling in the organizational designhas a second effect in a dynamic context. It affects notonly how performance varies in the face of environmen-tal change, but also the ability of the organization to learnand improve. Learning in an organization involves threesomewhat distinct processes. First, there must be variation,the identification of new alternatives. This can come fromexperimentation in the organization itself or from observ-ing what other organizations are doing. Internal experi-mentation might occur in the laboratory or productdevelopment if it involves new or improved products, butit could just come about from the actions of differentpeople and units in the business trying to solve emerg-ing problems or acting on new insights. Then there is theprocess of selection, of determining whether the newalternatives are indeed better than the current way.Finally, if something better has been identified, the newways must be transferred and retained in the organization.

Obviously, the more loosely coupled the design, theeasier it is to experiment with changes in any aspect of the

Key Concepts

71

organization. Indeed, with different parts of the businessfacing different opportunities and challenges and operat-ing in different environments, variation will come auto-matically if the units are sufficiently free from mandatedcentral control of their architectures and routines. Thus,loose coupling supports variation.

However, loose coupling makes selection and transfermore difficult. It may be hard to determine whether some-thing new that works in one part of the organization willalso work in another part that has other routines and per-haps a different culture. Moreover, if the different parts ofthe organization are each allowed to do what they thinkbest, then getting them to adopt a new practice or productmay be much harder than if the center can simply orderthat currently defined best practice must be followed.

Thus, there is a second trade-off in determining theideal amount of tight versus loose coupling. Not only isthere a trade-off between getting optimal performance in aparticular context versus doing better in the face ofchanges in the strategy or environment, but there is also atrade-off in supporting learning. A company that mandatesoperating strictly according to currently defined best prac-tice and has operations manuals that are always followedwill have a very hard time generating potential improve-ments in that best practice. One that allows more variationwill rarely be using best practice at any given point in timein all its operations, but may do better on average.

So we now have a way of thinking about strategy andorganization in the modern firm. Success involves strategyand organization that are coherent, fitting with one anotherand with the environment. However, environmental changeimplies a need to change strategy and structure—there is

Key Concepts

72

no one universal answer to the design problem. We under-stand that change will be difficult, and some reasons whythis is so. But we also understand that the choices are amongcoherent systems, not individual policies and features. Thisaccounts for the fact that the changes that firms have beenadopting in recent decades have definite patterns—they arenot random, but rather consist of general moves towardsa new coherent pattern of organizational design.

Notes

1. Hidden in this definition is the requirement that doingmore of one activity does not automatically prevent doingmore of the other. Note that complementarity is conceptu-ally different from a positive spillover. A positive spilloveroccurs when the overall benefit from some activity (ratherthan the returns to increasing the activity) is increasing inthe level of the other activity.

2. See Brynjolfsson and Hitt (2000) for a survey and an exten-sive bibliography.

3. The rates are intended to be set so that a worker producing atthe standard rate would earn an amount that would be com-petitive with standard industrial wages in the geographic area.

4. These papers also develop much of the theory of complemen-tarity more formally. The 1990 paper contains formal mathe-matical statements and derivations of the key results, while the1995 paper provides a more accessible verbal exposition.

5. For a formal model of these ideas, see Nickerson andZenger (2003).

6. For a fuller discussion, see Milgrom and Roberts (1994).A more easily accessible source for some of the argument isMilgrom and Roberts (1992: 349–52).

Key Concepts

73

74

3

The Nature and Purpose ofthe Firm

Why are there firms? What is their fundamental natureand purpose? The answers to these seemingly very aca-demic questions are actually of practical significance. Ourobjective is to understand the problem of designing thebusiness enterprise and to gain some insights into some ofthe changes we see in the way firms are being organizedand managed. Having a clear understanding of the natureand purpose of the firm as an institution is fundamentalto reaching the desired objective.

The primary answer is that firms exist to coordinate andmotivate people’s economic activity. As Adam Smith(1776/1937) noted in his famous discussion of a pin factory,scale economies and learning effects imply that there aretremendous gains in efficiency if individuals specialize intheir production activities (the division of labor). But oncepeople specialize, they become mutually dependent,because no one produces by herself all the things she needseven to survive, let alone to prosper. Indeed, in most contextsin a modern economy, an individual in her job actually produces nothing that she wants personally to consume.

Nature and Purpose of the Firm

75

Instead, she must exchange the limited set of things shedoes produce for the vast variety of goods and services thatshe actually wants and that others make. These interdepen-dencies mean that there is a need to coordinate differentindividuals’ activities and to motivate them.

Coordination means, at a minimum, that all the neededtasks are completed without pointless duplication. Betteryet, it seeks to ensure that the tasks are done efficiently, bythe right people, in the right way, and at the right timeand place. Ultimately, full coordination also requires thatthe tasks actually undertaken are the right ones. In thecontext of the firm, this means that the activities indi-cated by the strategy are carried out in a cost-minimizingfashion and that execution on the strategy creates asmuch value as possible. Finding a solution to the coordi-nation problem is clearly a major task, even in relativelysimple contexts. The coordination problem for the econ-omy as a whole is mind-bogglingly complex.

Motivation becomes a problem too, because it may notautomatically be in the self-interest of individuals orgroups to act in ways that promote realizing an efficientsolution to the coordination problem. Generally, wemight expect that people are somewhat selfish. This is notto deny elements of altruism, but just to assert that purealtruism is unlikely. Most people most of the time wouldlike to receive more of the things they value, even if in sodoing they might deny others the benefits of these things.Further, they will want to avoid as much of the costs ofeconomic activity as they can, even if others have to bearsomewhat higher costs as a result. In the presence ofinterdependencies, individuals’ attempts to grab more ofthe benefits and duck the costs can make everyone,


76

including themselves, much worse off than if theybehaved differently. The issue is then to motivate peopleso they choose to behave in ways that are conducive torealizing a coordinated solution.

Of course, as Smith also noted, markets are one veryprominent mechanism for solving the problems of coor-dination and motivation that arise with the interdepen-dencies of specialization and the division of labor. Marketinstitutions leave individuals to pursue (narrowly) self-interested behavior, but guide their choices by the pricesthey pay and receive.

A well-functioning market leads the interdependenciesamong people to be fully “internalized.” Interdependencemeans that one person’s choices and actions have an impacton other people. Selfish behavior would then potentiallylead to inefficiency, because the decision-maker takesaccount of only some of the costs and benefits of his actions,namely, those he experiences personally. With well func-tioning markets, however, each person is led to take fullaccount of the aggregate costs and benefits of his actions, nomatter to whom these accrue, because these are reflected inthe prices that he faces. Prices in well-functioning marketssimultaneously reflect the benefits of an extra unit to buyersand its cost to sellers; so price-guided choices bring thesemarginal costs and benefits into equality (as they must befor efficiency). In essence, market prices signal what needsto be done, when, where, how, and by whom. In so doing,markets achieve a remarkable level of coordination withoutany conscious central planning or control.

In fact, one of the central results of economic theory isthe demonstration that if all the relevant markets exist andare competitive, then the allocation of resources that

emerges from full market clearing is actually an efficientone—there is no rearrangement of economic activity thatwould be unanimously preferred. Instead, any change fromwhat the market generates must hurt at least one person.

Moreover, markets provide intense individual incentivesfor innovation, investments, and effort; they require min-imal amounts of formal communication about opportuni-ties, needs, and resources; and they allow for unmatchedindividual freedom and personal discretion. The case forusing markets to coordinate and motivate is strong.

Further, as Nobel laureate Ronald Coase (1960) hasargued, even if formal, organized, competitive markets donot exist, direct bargaining among the interdependentparties can produce the same efficient outcomes, againwithout conscious central planning and control and againwith each pursuing only his own self-interest. The logic issimple. Suppose that all the affected parties can meet andbargain freely and that the consequences of failure to agreeare clearly established (e.g. through freely enforceable prop-erty rights). Then the different parties have every reason toreach an agreement that is efficient in the sense usedabove—one where it is impossible to make one of the partiesbetter off without hurting one of the others. Otherwise,there are gains to be had from revising the bargain.

If voluntary dealings between distinct parties work sowell, however, why then do we use firms to coordinate andmotivate economic activity, particularly as extensively as wedo? After all, as Nobel laureate Herbert Simon (1991) hasnoted, even in the most market-oriented of economies, thevast bulk of economic activity occurs within formal, man-aged organizations rather than through market exchanges.In fact, John McMillan (2002: 168–9) estimates that less


77

than a third of all the transactions in the U.S. economy occurthrough markets, and instead over 70 percent are withinfirms. Coordination and motivation are predominantlyachieved through firms, rather than the price system orbargaining among individuals on their own account. Why?

The basis for the answer was put forward by yetanother Nobel laureate, Kenneth Arrow (1974). Arrow’sanswer is that sometimes markets simply do not work:There is “market failure.” It can happen that markets failto exist, or are not competitive, or do not clear properly.In these circumstances, interdependencies among peoplewill not be fully internalized—people’s self-interestedbehavior does not account for the costs borne and bene-fits received by others. Consequently, possible changes inthe allocation of resources that would benefit all partiesmay remain unrealized. When markets fail to yield anefficient solution to the problems of coordination andmotivation, other mechanisms for coordinating and moti-vating may be better and may come to supplant the mar-ket. The firm is the principal such alternative.1

By this logic, economic activity should occur withinthe firm when it represents a better way to coordinate andmotivate than does the market. To understand when thisis likely to be the case, we need to understand the natureof market failure as well as when firms might be expectedto work better.

Sources and Nature of Market Failure2

Microeconomics identifies a number of circumstanceswhen market failure is likely and thus other arrangements


78

may be preferred. The most familiar include situationswhere monopoly or other forms of imperfect competitionprevail, whether through collusion or because barriers toentry or regulation limit the number of competitors. Insuch circumstances, supply is typically restricted toincrease profits. This hurts efficiency because customers’losses in higher prices and foregone surplus typicallyexceed the extra profits that are generated for the monop-olist. If internal provision of the monopolized good is rea-sonably cost efficient, then internal procurement may bebetter for the customer.

Public goods are another traditional example wheremarkets do not work well. A public good is one where oneparty’s consuming the good does not diminish the amountavailable to others; indeed, in the extreme, making the goodavailable to one person may even require that it be availableto all. The first property is called nonrivalry in consump-tion; the second is called non-excludability. An example ofa non-excludable public good is national defense—if oneperson in an area receives it, all do. Television broadcastsare an example of an excludable public good. One person’sviewing the signal does not diminish its availability toothers, so consumption is not rivalrous, but the signal canbe scrambled to prevent its being viewed by those who donot have a decoding device. Simple market arrangementsdo not work well for public goods because one person’spurchases are automatically available to all (in the non-excludable case), and there will then be a temptation to freeride on others’ contributions. Alternately, the individualbenefits that will drive someone’s decision to purchase orprovide a public good are only a fraction of the aggregatebenefits, leading to undersupply. The inefficiency with


79

excludable public goods is that, in order to induce people topay for the good, resources are expended to deprive non-payers of the benefits. Yet, once the good has been created,these extra benefits could be made available to all at no costto anyone, increasing the total benefits realized from thegood at no direct cost.

Since information often has the characteristics of anexcludable public good, it is to be expected that marketsfor information are likely to be problematic. In fact, theproblems of selling information are especially intense.How is the buyer to evaluate the value of the informationbeing offered? With a private good she can examine it andtry it out. But if the seller of information offers theseoptions, the buyer has gotten the value for nothing,because the information cannot be taken back on herrefusal to pay. Sometimes the information can be codifiedand packaged in a way that makes it more easily exclud-able. For example, potential customers can be allowed totry a database or computer program and if they do notagree to pay, then the sample can be taken back. At othertimes a patent may be created and enforced. But generallyproblems remain and inefficiencies can persist.

Public goods are an extreme form of externality. Anexternality exists when a person’s actions affect others’welfare and the first party does not have an incentive torecognize this impact in decision-making, so that he or shedoes not account for all the costs and benefits in selectingactions. Because of this, externalities lead to inefficiencies.

The classic examples have to do with such choices as adecision to drive to work on congested roads using a pol-luting vehicle, but externalities arise in business too. Forexample, there is a danger that different business units


80

will ill-use a common corporate asset that is available toall, such as a common brand. Because the benefits of pro-tecting the brand are spread among all those that use it,whereas the costs of taking actions to protect and build itare borne by the individual units, there will be inadequateincentives to invest in brand building and maintenance.

Coase’s argument that direct bargaining can replaceimpersonal market dealings to achieve efficiency wasformulated explicitly to deal with externalities—he citesexamples such as soot from a train raising the costs for alaundry that dried clothing outdoors or sparks from thetrain causing fires (Coase 1960). Coase argued that if prop-erty rights were well established and enforceable, then bar-gaining would lead to an efficient outcome. For example, ifthe train owner has a property right to pollute but it ischeaper to reduce the emissions than for the laundry torewash soiled items, then the laundry will pay the trainowner to reduce the pollution. A similar point applies tothe example of monopoly: There is no obvious reason for amonopolist to adopt value-destroying behavior. Instead, letit bargain with potential customers. If Coase’s argumentsapply, then the parties will reach an efficient outcome.

The issue then is why such bargaining might not work.Part of an answer lies simply in the costs of identifyingthe relevant parties, bringing them together to negotiate,establishing the terms, and then enforcing the agreement.Many of these are, in turn, tied up in information prob-lems. In fact, recent research has given much attention toinformational problems as a source of market or bargain-ing failure, and the originators of this work, JamesMirrlees, George Akerlof, Michael Spence, and JosephStiglitz, received Nobel prizes for their contributions.


81

One especially important class of circumstancesgenerating market failure arises when there are informa-tional asymmetries among different parties. For example,suppose potential sellers are better informed than thebuyers are about the quality of the items they are offering.Think, for example, of used cars: The original owner’sexperience with the vehicle makes him much betterinformed about the car’s quality than is any potentialbuyer. Then the buyers must be wary that they are notbeing stuck with a really low quality car—a “lemon”—ata high price, because sellers with low quality items may betempted to misrepresent the quality of their wares. Thus,buyers will expend resources to determine the quality ofthe goods they are offered, and sellers of higher-qualitygoods will attempt to demonstrate that they really do havegood products that will command premium prices (Spence1973). Each of these activities involves costs but directlycreates no value and so results in a less-than-efficientallocation.

The inefficiency can be even more substantial thansimply the waste of resources in screening and signalingquality: Trade may break down almost completely(Akerlof 1970). If eliminating the asymmetry of informa-tion is not possible, then buyers will refuse to pay morethan the expected value of the goods, averaged across thedifferent quality levels they expect to be offered. Thenthe best quality goods may not be offered at all, becausethey command only a middling price that does notreflect their true value. Consequently, the distribution ofqualities that are actually offered is worse than what ispotentially available. Since the selection of products onoffer is not representative of the underlying distribution


82

of quality, but instead is an adverse selection,3 buyers willrationally lower their willingness to pay even further.Then, even more potential sellers of relatively high-quality items may no longer be willing to sell at the lowerprice. The overall result may be that nothing but very lowquality items are available—only lemons are on offer—and markets fail to exist for high-quality products,although buyers are anxious to have such goods andwould happily pay enough for them to compensate thesellers if they were sure to get what they paid for.

A second form of informational asymmetry that maycause market failure involves limitations on the ability toobserve others’ actions, and thus to determine whetherthey are adhering to agreements. This leads to the prob-lem of moral hazard.4 Is a salesperson in the field actuallycalling on customers or playing golf? Is a knowledgeworker really thinking about the job, or about somethingelse? Is my lawyer giving me an honest best effort or aperfunctory, half-hearted one? Similar problems arisewhen the information that individuals have gathered andupon which they are acting is not observable by others inwhose interests the actions are to be taken. Is my broker’srecommendation to trade really a good one, or does hejust want the commissions? Am I being denied a possibletreatment for my medical condition because it really willnot be worthwhile or because the insurance companydoes not want to pay for it?

In such circumstances parties cannot write contractsthat directly address their concerns, and markets for thedirectly relevant goods and services then cannot exist.Instead, parties must content themselves with mea-suring performance indirectly and imprecisely and with


83

contracting on proxies and signals rather than directly onthe value-creating choices and actions. This can createvarious sorts of inefficiencies: Misdirected effort, a mis-allocation of risk, inefficiently low levels of effort, expen-ditures on monitoring and on manipulating theperformance measures, and so on. (We will deal with theproblem of moral hazard in detail in Chapter 4.)

Bringing transactions that are subject to adverse selec-tion and moral hazard out of the market and under someother form of organization does not automatically solvethe problems. However, the fact that market dealings donot work well does mean that other arrangements may dobetter than the market and are worth exploring.

Much recent research has focused on another problemof enforcing agreements. Even if there are no informa-tional asymmetries between the directly affected partiesthat prevent contracting on the relevant variables, relyingon third parties to enforce a contract requires that theseparties must be able to determine whether the agreementhas been breached. It is easy to imagine situations wheresuch external verifiability might be problematic: Bothparties to an agreement might be fully aware of what eachhas actually done, and yet there is no unambiguous way ofdemonstrating the facts to others. In such cases, there aretwo options. Either useful agreements cannot be reachedregarding the nonverifiable matters, or else the agree-ments must be “self-enforced”—the parties must findit in their interests to adhere to the agreement, even inthe absence of the incentives that might normally comethough outside enforcement. Simple market-likearrangements can rarely provide such self-enforcementoptions.


84

Somewhat related are problems of commitment. A simpleinstance arises when contract enforcement is costly.These costs might be legal fees, court costs, and man-agers’ time, which in some cases might exceed the gainsthat could be realized by winning. If the parties to anagreement do not have an incentive to insist on itsenforcement, contracts again may lose their efficacy. Ifenforcement is costly, then both sides may lose from pun-ishing transgressions. Absent some means for theaggrieved party to commit to punishing the transgressor,some violations of the agreement will then go unpun-ished. They consequently cannot be prevented. This inturn affects what can be achieved by contract.5

A subtler but also more important commitment prob-lem arises when actions taken under the agreement alterthe incentives to adhere to the remaining terms. This mayresult in a situation where both parties would be happy torenegotiate. Yet, understanding that the renegotiationmay occur will affect the incentives to abide by the origi-nal agreement. For example, suppose that to motivate anemployee, her pay is made very highly dependent on herperformance. Suppose too that performance is not com-pletely under the employee’s control but instead is alsosubject to random, uncontrollable variation. Once the jobhas been done, but before the results are realized, there isno further value to having the employee bear all the riskthat making her pay depend on her performance entails.At this point there is nothing more to motivate, and yether continuing to face risk in her pay is costly if she isrisk-averse and the firm is not (as might reasonably be thecase). Both the employee and the firm would gain byreplacing the original payment scheme with a fixed


85

payment that was somewhat lower than the expectedvalue of the original reward. The firm saves money inexpectation and the employee avoids the risk. Yet, if theemployee forecasts that renegotiation of this sort is likely,she may no longer be motivated to exert effort in the firstplace.

While both sides are happy to renegotiate in this exam-ple, there can also be situations where the renegotiationessentially occurs under duress. For example, supposethat two companies enter a supply relationship in thecourse of which the buyer becomes locked into dealingwith the seller because no competing supplier can meetthe buyer’s needs. Then the seller may be able to use itspower to hold up the buyer, forcing a renegotiation of theoriginal contract terms to deprive the buyer of many ofthe benefits it forecast when the relationship began. Thisbargaining may be costly, and it may lead to a breakdownin the relationship. Moreover, foreseeing the possibility ofopportunistic renegotiation may prevent the relationshipever being formed (Williamson 1975; Klein, Crawford,and Alchian 1978).

The source of difficulties here is that it is not possibleto commit not to renegotiate. Why should there be suchcommitment problems? In the first instance, the courts(at least in the United States) generally will not enforce a“no renegotiation” clause in a contract—if the partiesagree to renegotiate, they are free to do so, no matter whatthey may have previously agreed. There is some questionabout the desirability of such a policy, but even in itsabsence, preventing renegotiation may be difficult. Inmany cases, there is no third party that would have aninterest in enforcing the prior agreement, and neither


86

contracting party would want to do so if the renegotiationis mutually beneficial.

Moreover, since any real contract is necessarily incom-plete, there will very likely be situations that arise that werenot foreseen in the original contract’s terms. These mayrequire some sort of after-the-fact negotiation. Contractualincompleteness can arise for a variety of reasons. If actionsor outcomes are not verifiable by third-party enforcers,there is no point to contracting on them explicitly.Bounded rationality can prevent parties from foreseeing allthe relevant contingencies that might arise. The costs ofnegotiating terms of agreement concerning contingenciesthat are not thought very likely to occur can rationally leadto their omission from the contract. The inherent limita-tions of natural language may prevent describing termsunambiguously. In any of these circumstances, after-the-fact renegotiation may be needed in particular situa-tions. This means that there is a cost to applying bans onrenegotiation unconditionally. Meanwhile, distinguishingsituations where renegotiation is to be allowed and where itis forbidden could be extremely problematic.6

Generally, if contracts are ineffective or markets simplyfail to exist, they clearly cannot guide efficient resourceallocation. Coase’s bargaining might be a solution in somecircumstances where organized markets do not come intobeing. Yet, the same informational asymmetries thatundercut the working of the market would plague bar-gaining, preventing it from reaching fully efficient agree-ments.7 Meanwhile, in many of the cases of imperfectcompetition or public goods, the costs of organizing allthe relevant parties to bargain and of enforcing any resul-tant agreements might be overwhelming.


87

When markets do not work, other institutions may becreated that do a better job. In particular, the firm can besuch a mechanism.

Firms versus Markets

When might firms be better than markets? Much of econ-omists’ current understanding of the answer to this ques-tion traces to another element of the work of RonaldCoase. Almost 70 years ago, Coase (1937) asked explicitlywhy some economic activity is carried out through markettransactions while other parts are organized under hierar-chic authority relations within firms. His answer is thatthere are costs to organizing economic activity, to achievingcoordination and motivation, and that economizing onthese transaction costs explains the patterns of organiza-tion that are adopted. In particular, a transaction isremoved from the arm’s length contracting of the marketand brought inside the firm precisely when it is cheaperto organize it this way. Thus, we are to understand theboundaries of the firm and, more generally, observed pat-terns of organizational design as being efficient ones—ones that create the most possible value.

There are at least two aspects of Coase’s answer thatneed elaboration. One is why efficiency—rather than,say, the pursuit of monopoly power and profits—shouldbe determinative. The second is the origin and nature oftransaction costs.8

The basis for efficiency arguments is simply that ifarrangements are not efficient, then (definitionally) it ispossible to make everyone better off—not just to increase


88

the size of the total pie, but instead actually to increasethe size of each person’s slice. Presuming that the poten-tial improvements can be identified and the gains can beshared, we should expect such changes to be made. So ifan arrangement persists, there is reason to suspect that itis efficient, at least for the parties who are in a position tohave their interests represented.9

This argument might seem Panglossian—that every-thing we observe is the best it could be—but there areimportant qualifications and subtleties that render it lessproblematic.

Note first that the requirement that the relevant partiescan identify any potential improvements limits what isachievable. Thus, the efficiency of actual arrangements isconstrained by informational and observational limitations.

An example is the “Market for Lemons” problem ofadverse selection considered earlier. The result of theinformational asymmetry may be that only the very worstcars are offered and sold, even though there are manypotential trades that would make both sides to the trans-action better off. Gains from trade are not realizedbecause informational asymmetries prevent identifyingtheir magnitude and sharing them in a satisfactory way.So this situation may, in fact, be the best that can beachieved (provided we continue to respect private prop-erty and allow each side to decide whether it wants totrade). Thus, it is efficient in the limited sense that we usethe term, although this mainly demonstrates how weak anotion efficiency actually is, or just how constraining theinformational limitations are.

Similarly, strikes and other costly delays in reachingagreements need not be seen as inefficient waste and


89

evidence against the efficiency hypothesis. Rather, theymay be interpreted as the best options available under thecircumstances for credibly communicating the value of anagreement to each side (Kennan and Wilson 1993). Forexample, a firm’s willingness to suffer a strike signals thatan agreement is not so valuable to it as might have beenbelieved. Thus, the union learns that it cannot hope for asrich a settlement as it had desired. For if an agreementwere really very important and valuable to the firm, itwould be anxious to settle.

We now turn to the second point: the nature of trans-action costs. In a market setting, transaction costs are thecosts of finding and qualifying trading partners, of estab-lishing specifications and prices, of negotiating anddrafting contracts, and of monitoring and enforcingagreements. They are also the opportunity costs of lostbenefits that are occasioned by the difficulties of develop-ing complete, enforceable agreements between separateparties.

To a large extent, the informational and commitmentissues discussed already underlie the transaction costs ofusing markets. However, one particular example has a cen-tral place in the research in this area. The exampleinvolves hold-up and specialized investments (Williamson1975, 1985; Klein, Crawford, and Alchian 1978).

Williamson (1975) argues that many business dealingsinvolve lock-in—even if there are initially lots of potentialtrading partners, once one is chosen and the parties startto work together, there is a “fundamental transformation”of the relationship that makes changing to another partnervery difficult. In such circumstances, if contracts areincomplete, then the parties may have to negotiate after


90

the lock-in has occurred. These negotiations may be costlyand acrimonious in the best of circumstances. Moreover,they present an opportunity for one party to act oppor-tunistically to attempt to extract more of the returns tocooperation than it was due under the original agreement.Both the bargaining costs and potential benefits that arelost if the bargaining breaks down and cooperation doesnot occur are transactions costs of dealing with anotherparty.

Lock-in is actually inevitable when assets are special-ized. An asset is specialized to a particular use when thevalue it can create in its next-best alternative use is sub-stantially lower than what it yields in the current one. Forexample, the dies used to shape materials in manufactur-ing are very specific to that use: If they are not employedfor this purpose, they are just scrap metal. Firm-specifichuman capital—knowledge that is only (or especially)valuable in the context of employment with a particularfirm—is another example. When assets are specialized,they are subject to hold-up—attempts by trading partnersto appropriate some of the returns that the assets’ ownersexpected when they invested in them. This can lead to avariety of inefficiencies.

Suppose two firms have an opportunity to trade, butthe seller needs to make specific investments to serve thebuyer’s needs in the best way possible. Once the invest-ments are made, the costs are sunk. This means that evenif the price ultimately received by the seller were cutalmost to the level of variable costs, so that almost no con-tribution to covering the costs of the investments wouldbe realized, it would still not be worthwhile to withdrawfrom serving the buyer. The reason is that the sunk costs


91

must be borne in either event and the asset has no othergood use. A portion of the returns to the asset are thenquasirents, returns in excess of what is needed to keep theasset in its current use once it has been created.10 Theseller is then subject to the danger of hold-up.

If there is a prior contract, but it is sufficiently incom-plete that negotiations over terms need to take place afterthe investments have been made, then the bargaining overterms will very likely give little protection to the seller’sinvestments. This is because the sunk costs are irrelevantin determining how much value is created by cooperationversus breaking off the relationship, which is what theparties effectively bargain over. Even if the terms arenominally fixed in advance, the buyer may still be temptedto force a renegotiation of the terms of trade, appropriat-ing a portion of the quasirents that the seller had hopedto enjoy. This is possible because the seller has littlerecourse: To refuse to renegotiate and break off the dealleaves him or her with only the nearly worthless asset.Meanwhile, forcing a renegotiation may be quite simple.For example, the buyer could claim business conditionshave changed in way that justifies a lower price, or thatservice or quality has not been acceptable, or any ofa number of other things, depending on the particularcircumstances.

Thus, the seller cannot expect to receive the full returnson the specific investments it has made. Anticipating this,the setter may be reluctant to commit resources to thespecific assets. For example, if the specificity arises fromthe seller’s learning the particular needs of the buyer, theseller might underinvest in this knowledge, so that less ofa loss is suffered in case of a hold-up. Thus, less value is


92

created. Alternatively, the seller may expend resources forprotection against the anticipated hold-up. Making theassets more flexible in their uses, so that they can be rede-ployed at less cost, might do this. This is a waste, for theresources are being expended to improve the value of theasset in a use to which it ought not to be put.11

One solution to this problem is for the buyer to pay a partof the cost of the investment up front—essentially thebuyer pays ex ante for the amount to be (mis)appropriatedlater. This will work, however, only if the agreement toundertake the investment is enforceable. Otherwise, forinstance, the seller might just pocket the buyer’s money andstill make only the investment that seems individually opti-mal given that the terms will later be renegotiated. Anothersolution may be for the transaction to be brought within asingle firm. Empirically, this has been an important elementin vertical integration.12 This is can be costly in a number ofways, however, as we will soon see. Thus, the enforceabilityproblems create transaction costs in markets.

What are the transaction costs of organizing economicactivities inside the firm? This is still a controversial issue.One might think first of the costs of communicatinginformation up and down through the hierarchy, of infor-mation overload at the center/top, and of slow decision-making that is based on limited and possibly outdatedinformation. Organizational decentralization may some-times provide an effective response to these phenomena,however, as the developers of the multidivisional formdiscovered (Chandler 1977). (Generally, it is a good ideanot to rely on explanations that are based on ineffi-ciency—managers are awfully good at creating new andbetter ways to do business more efficiently!).


93

In this vein, Oliver Williamson (1985) has pointed tothe policy of selective intervention as a response to anyinherent disabilities of the centralized, hierarchic organi-zation of a firm. The idea is to replicate the workings ofthe market within the firm whenever this yields efficiency,while top executives intervene selectively in the subunitsand the relations among them only when this yields abetter outcome than market dealings.

If selective intervention worked, then it would be effi-cient to have everything in one gigantic firm. Yet, even theideologues of the old Soviet economy never dreamt of asystem that was so extreme in its centralization. Theremust be something that prevents effective application ofselective intervention.

One response is that it is impossible to generate thesame intensity of incentives within a single integratedfirm as when units are separately owned. In this regard,Williamson himself suggested that, while it might be easyto promise as strong incentives to employees as to outsidecontractors, it is hard to do so credibly. The problem isthat the owner controls the performance measures13 andwould always be tempted to fudge them. This could hap-pen both when the employee has done very well and isdue to be paid a lot, and when there have been bad resultsdespite apparently good effort, in which case the ownermay be too forgiving. Either possibility blunts actualincentives and can imply that the firm does not achievethe levels of efficiency that the market might realize.

This argument clearly rests on the difficulties of effec-tive contracting. Reputational concerns may help counterit. As well, it may be possible in some circumstances touse third-party monitoring and auditing. For example,


94

BP used “self-help” figures—essentially, improvementsin earnings not resulting from changes in crude oil pricesor exchange rates—in its performance pay. It thenemployed an outside auditor to attest to the accuracy ofthe self-help numbers it calculated. Equity carve-outsand tracking stocks may be an especially interesting pos-sibility here. For example, Thermo-Electron Corporationsold stakes in its business units to the public explicitly to“outsource performance evaluation.” Outside equityinvestors are strongly motivated to act as monitorsbecause their own funds are on the line, and the stockprices they generate become low-cost, objective perfor-mance measures that may have more credibility andintegrity than any internally generated ones.

The “property rights” approach to the theory of thefirm, developed by Sanford Grossman and Oliver Hart(1986) and Hart and John Moore (1990),14 suggestsanother reason why it may be harder to give strong incen-tives in a larger, integrated organization. This logic is mostapplicable when thinking about owner-managed firms.Suppose such a firm is selling to an industrial customer. Ifthe relationship is severed, the owner of the upstream firmstill owns the assets in his firm (machines, brand name,etc.) and can redeploy them as he sees fit. In contrast, sup-pose the customer owns the assets, with the upstreammanager now an employee running a business unit corre-sponding to the original firm. Now if the relationshipcollapses, the manager does not get to keep the assets.

As Grossman, Hart, and Moore (GHM) argue, this dif-ference affects the relative bargaining position of the twoparties in dividing up the value created by their coopera-tion. (Assume that it is impossible to specify the division


95

of value contractually in advance, so that it must bedetermined by bargaining after the value has been real-ized.) Thus, the ownership of assets determines the pay-offs the parties receive.

These payoffs in turn affect the strength of the incen-tives the parties have to undertake investments that arecomplementary with the assets of the firm/business unit,such as learning how to work with the assets more effec-tively or developing a brand that increases the value of thegoods that the buyer produces using the services of theassets. Getting a larger share of the return motivatesinvesting more to create greater returns. So, who owns theassets affects investment and thus the value created. Ifthere are two separate firms interacting through the market,the supplier owns the assets and has strong incentives toinvest, but the buyer’s incentives are weak. If there is ver-tical integration and the buyer owns the assets, then theemployee–manager has weak investment incentives,although the buyer has strong ones. With incompletecontracts it simply is not possible to give the same incen-tives to an employee as an owner receives.15

Note the importance of incomplete contracts to thistheorizing. If binding agreements were possible, then thedivision of the value created could be set contractually toprovide incentives, or, indeed, the investments them-selves could be governed by contract. (In this, the GHMtheory is like the hold-up analysis discussed earlier.) Then equally strong incentives could be given inside thefirm as outside—ownership and the boundaries of thefirm would not matter.

Bengt Holmström and Paul Milgrom (1991) have arguedthat the issue is not just offering incentives that are strong


96

enough, but offering ones that are appropriately balanced.The full details are in their model of multi-tasking inagency relationships, which is discussed in detail in theChapter 4, but the theory rests on two observations. First,they note that typically there are multiple ways that some-one can spend time, many of which might be of value to anemployer. But if these activities compete for the person’sattention,16 then the incentives offered for different activi-ties must be comparable. Otherwise, the person will focusdisproportionate amounts of her effort on those things thatare especially well compensated and ignore the others. Thesecond observation is that providing strong financial incen-tives is costly if the person is risk-averse, because it loadsextra risk into pay. Further, the cost is greater the more dif-ficult it is to measure performance. This means that, otherthings being equal, tasks where performance is hard tomeasure should not be given as intense incentives as onesthat are more accurately observed.

Suppose now that two activities are desired. Think ofone as producing output, which is easily measured,implying that the costs of providing strong incentives (interms of the risk that the person bears) are low. In isola-tion, this activity should then be given strong incentives.The other can be thought of as some form of investment,where effort is hard to measure accurately and in a timelyfashion. For example, it is hard to determine precisely thechange in the long-term value of a division occasioned byits manager’s efforts and decisions. Providing strongincentives for this investment activity is very costly. Thisis because doing so makes her pay highly random, since itis not determined solely by the manager’s actions but alsoby the other uncontrolled factors that affect measured


97

performance. The manager will have to be compensatedfor bearing this risk, so the costs ultimately are borne bythe employer.

It is obviously desirable that the manager both increasecurrent performance and undertake the right investmentsthat increase long-term value. If, however, strong incen-tives are given for improving current costs and revenuesand weak ones for investments (as might be optimal ifthere were no interaction among tasks), then problemsarise. The manager is tempted to mortgage the future,ignoring good investments, and concentrate on gettingcurrent performance up, even if this reduces total valuecreated. The solution must be to provide balanced incen-tives. There are two ways to do this.

One solution is to sell the operation to the manager,who then bears the long-term consequences of her invest-ment choices as well as the current effects of improvingperformance. This may, in fact, be a factor in the man-agement buyouts that first became prominent in the1980s. The second is to treat the manager as a salariedemployee, giving relatively weak incentives for bothshort- and long-term performances. (These incentivesmight be implicit and subjective, perhaps through theopportunities for promotion.) The first solution meansthat the manager receives as strong incentives for gener-ating future returns as current ones. The second meansthat both sorts of effort get equally muted incentives. Ineither case, the incentives are appropriately balanced, andboth activities get some attention (but less, of course, inthe low-incentives, employment regime).

The key point for the present discussion, however, isthat if the employing firm continues to own the investment


98

opportunities, the employee must be given weak incentivesfor other activities—weaker than what would be received asowner of a separate firm. Thus, the market solution cannotbe replicated inside the firm.

A fourth approach to the issue of why selective interven-tion does not work questions whether senior managementwill—or whether they even can—limit their interventionsto those that are efficiency enhancing (see Milgrom andRoberts 1988b, 1990a, c, 1992: 192–4 and 269–77, 1998). Adefining characteristic of the firm is that its executives havethe unchallenged legal right to intervene in lower-leveloperations and decisions, to direct that very specific actionsbe taken, and to enforce these directives. (Indeed, theymust have this power if selective intervention is to occur.)In contrast, outsiders (even the courts or regulators) can-not easily make such detailed interventions. So moving anactivity out of the market and into the firm increases theopportunity for interventions, including ones that are notefficiency enhancing.

Excessive or inappropriate interventions might comefor a number of reasons. First, senior managers may betempted to intervene when they should not because, afterall, it is their job to manage. They may also be too impa-tient, intervening when they see that at lower levels people might not do the absolutely best thing. This isunderstandable—mistakes are being made, after all—butcostly. The intervention destroys both the opportunities andthe incentives for the lower levels to learn. It also undercutstheir autonomy and the very real performance incentivesthat come from that (Aghion and Tirole 1997). The seniormanagers can also have an overblown estimate of their own abilities, not trusting others to take the appropriate


99

actions (i.e. those that they would take themselves).Finally, implicit bribery of various forms might lead themto intervene.

But even if the executives are scrupulously honest andsuperbly competent, there may still be excessive interven-tions. The problem is that lower-level people will care alot about the decisions that the firm makes, and they willhave every reason to attempt to influence the executives tointervene, making the decisions in the way they like. Forexample, someone in the organization has to be assignedto Paris, Texas, and someone to Paris, France. One canimagine that candidates would exert huge amounts ofeffort to affect this decision. Similarly, one person will bepromoted and another not, or one division’s investmentswill be funded and another’s not. There will be strongincentives for the interested parties to try to influencethese decisions, and not necessarily in directions thatincrease overall value creation. Further, to know whenand how to intervene, the senior executives will have torely on information from the potentially affected parties.

Among the techniques of influence are biasing infor-mation provision, misdirecting effort (e.g. towards build-ing the case for your side rather than attending to ongoingresponsibilities), politicking, and worse. Collectively,these influence activities have three sorts of costs. First,resources are directly expended on influencing decisions,even when all that is accomplished is to shift their distri-butional consequences (which means that no extra valueis created). Note that such efforts will also call forthdefensive expenditures from those who are threatened.The second is that, to the extent that the influence activ-ities are successful, bad decisions may be made. The third


100

is that the firm may be led to change its organizationaldesign from what would otherwise be ideal in order tocontrol the influence activities.

There are a variety of methods that can be employed tolimit influence activities. One is to limit communicationbetween executives and lower-level people. This limits theopportunity for politicking and strategic informationprovision, although it brings an obvious cost that someuseful information is not transmitted. The “three strikesand you’re out” policy employed at ABB Asea BrownBoveri, the Swiss–Swedish electrical power equipmentand industrial products company, is of this sort (Bartlett1993). Two managers who disagreed could take theirissues to a higher level for resolution, but only twice. Ifthey did it a third time, one or both was replaced.

A second approach is to structure decision processes sothey are less susceptible to influence. Firm adherence tobureaucratic, inflexible rules can be an example. If salariesare completely determined by seniority and job assign-ment, then there is no point in politicking for a raise. Theairlines’ policy of assigning cabin crews on the basis ofseniority similarly minimizes influence opportunities.Promoting on objective measures of past performancerather than on the apparent (i.e. less objective and moremanipulable) qualifications for the new post can be ratio-nalized as reducing the incentives for influence (as well asmotivating current performance). A very lean headquar-ters is one means to commit not to intervene too often—HQ simply lacks the resources to mess around in the lowerlevels’ business. Executives may also attempt to establisha reputation for not intervening in order to deter attemptsat inducing interventions. Note, however, that this may


101

require not intervening even when it might seemappropriate in a particular instance.

A third approach is to limit the distributionalconsequences of decisions, so individuals will have less atstake. For example, pay compression and uniformity oftreatment, even when other factors argue for differentia-tion, have this effect. This may explain the pressures thatare common in organizations to apply standard proce-dures even when differentiation and special treatmentmight seem appropriate—to do otherwise is to inviteeveryone to try to make the case that he qualifies as anexception.

A final method of controlling influence takes advantageof the fact that the boundaries of the firm set limits to suchinternal influence activities. Putting activities in separatefirms limits influence activities. For example, different payand promotion policies can be employed in connectionwith the different activities if they are in different compa-nies, whereas differential treatment within a single firmmight lead to huge amounts of politicking and influenceactivities. The reason the boundaries of the firm matter isthat it is pointless to campaign with one’s boss over assign-ment to another firm, but it may be quite reasonable to doso if the transfer is within the corporation. Similarly, oneadvantage of using outside suppliers rather than in-houseones is that trying to discipline or replace the in-housesupplier for poor performance invites influence costs, asdo the establishment and adjustment of transfer prices.

Once it is established that there are costs to internalorganization, the boundaries of the firm are deter-mined by the Coasian formula: Organize transactionsinternally if and only if the costs of doing so are lower


102

than organizing through markets. We know quite a bitabout market organization. How then to think aboutinternal organization? What characterizes a firm? Whyand how are firms different from markets? When will thefirm be the favored form?

The Nature of the Firm

Many authors, including Ronald Coase (1937) andHerbert Simon (1951), have identified the essentialnature of the firm as the reliance on hierarchic, authorityrelations to replace the inherent equality among partici-pants that marks market dealings. When you join a firm,you accept the right of the executives and their delegatesto direct your behavior, at least over a more-or-less com-monly understood range of activities. Simon argued thatthis may be an efficient response to the impossibility offoreseeing and contracting on what tasks will need to beundertaken—the need for coordination—and to theimpossibility (high costs) of bargaining anew each timethere is a change in the required activities. It is not a per-fect solution, because the boss will not have an automaticincentive to take account of the employees’ interests inchoosing how they spend their time. Still, it can be betterthan a rigid prespecification of activities, as under a sim-ple market contract.

Others—most notably Armen Alchian and HaroldDemsetz (1972) and Michael Jensen and WilliamMeckling (1976)—have challenged this view. They arguethat any appearance of authority in the firm is illusionary.For them, the relationship between employer and


103

employee is completely parallel to that between customerand butcher. In each case, the buyer (of labor services ormeat) can tell the seller what is wanted on a particular day,and the seller can acquiesce and be paid, or refuse and befired. For these scholars, the firm is simply a “nexus ofcontracts”—a particularly dense collection of the sort ofarrangements that characterize markets.

While there are several objections to this argument, wefocus on one. It is that, when a customer “fires” a butcher,the butcher keeps the inventory, tools, shop, and othercustomers she had previously. When an employee leaves afirm, in contrast, she is typically denied access to thefirm’s resources. The employee cannot conduct businessusing the firm’s name; she cannot use its machines orpatents; and she probably has limited access to the peopleand networks in the firm, certainly for commercialpurposes and perhaps even socially.

The firm’s control over access to resources comes ulti-mately from ownership of the assets and from being theunique common party to all the employment contractswith the members of the firm. By controlling access, thefirm can offer or deny people the opportunity to createvalue and earn rewards by being part of the firm. Thisgives the firm power, and it uses the power to specify the“rules of the game”: To prescribe and proscribe behavior,to set rewards and punishments, and to control relationsamong members of the firm and with outsiders. Thepoint of doing so is to create value by making the firm aneffective mechanism for coordination and motivation—a more effective mechanism than simple market relations.

These ideas (due to Bengt Holmström 1999—see alsoRajan and Zingales 1998) bring us back to Arrow’s


104

conception of the firm as a mechanism for dealing withmarket failures. This is actually a somewhat more gen-eral conception than the transaction costs approach,because we can think of the market as effectively failingwhen the costs of using it are higher than the costs ofnon-market organization. Note too that the central rolefor power harks back to the idea in the property rightsliterature that ownership is important because it conveyspower.

One striking insight from this line of reasoning is thatit is not necessarily a disability of the firm that it offersweaker incentives. Rather, one reason the firm exists isspecifically to provide weak incentives when those themarket provides are too strong!

What does it mean for incentives to be “too strong”?This is best understood in the context of theHolmström–Milgrom multi-tasking model discussedearlier: Incentives for some activity are too strong if theycause excessive diversion of effort and attention fromother valued activities that, for whatever reason, cannotbe given similarly strong incentives.

As an example, consider the question of whether to sellthrough an outside distributor or use an in-house salesforce. Suppose that, in addition to generating currentsales, the sales people (whether employees or not) can alsogather customer information that could, for example, beuseful in product development. If an outside distributoris used, it must be offered strong incentives for currentsales, for example, via a large commission that is perhapsthe full difference between wholesale and retail prices (aswith a independent retailer). Otherwise it will be inclinedto divert its attention to selling other clients’ products.


105

This means that the outside distributor must also getintense incentives to gather and transmit information if itis to be motivated to do this as well. Providing such incen-tives is likely to be difficult, however: How do you mea-sure performance in this activity?

Thus, if information gathering is important, so thatbalanced incentives are crucial, it may be better to giveweak incentives for sales that match the weak incentivesthat can be given for information gathering. But this can-not be done except by bringing the sales activity insidethe firm, where the sales force can be paid a salary, askedto gather information as well as to sell the product, andtold not to sell other firms’ products (Anderson 1985;Anderson and Schmittlein 1984; Holmström andMilgrom 1991).

Cooperation and Initiative

More generally, Holmström has pointed to a fundamentalmulti-tasking problem and associated set of trade-offsthat are involved in organizational design. Generally, twobroad sorts of behavior might be desired from people inan organization: Call them “initiative” and “coopera-tion.” The former refers to intelligent, honest, diligent,imaginative pursuit of individual goals and responsi-bilities—increasing your unit’s sales, lowering costs, mak-ing successful product innovations, and so on.Cooperation refers to promoting others’ well-being andcommon goals—improving another unit’s profits, devel-oping the overall brand, creditworthiness, and customerreputation of the firm, and so on. Clearly, both sorts of


106

behavior are desirable. Initiative leads to better individualand unit performance, both immediate and longer term.It is obviously valuable. But as long as there are any exter-nalities, cooperation is needed too, so that the interactionsare accounted for and managed. For example, we want toavoid hold-ups of other parts of the company, cheating onunobserved quality on internal sales, and depreciating thevalue of common assets and to encourage sharing knowl-edge and helping other members of the firm.

The problem is that we are in a multi-tasking context.Providing incentives to induce more of one sort of behav-ior may then involve getting less of the other. For exam-ple, if skimping on quality increases my profits but hurtsyours more, then initiative drives me to skimp, whilecooperation means I do not. The two are incompatible.

Typically, for any given level of expenditure ofresources on coordination and motivation, there will be amaximum amount of initiative that can be induced for anygiven amount of cooperation that is desired. Further, wemight expect that in most situations the frontier is down-ward-sloping: Once we have induced the most initiativepossible for a given level of cooperation, getting even moreinitiative is possible only by lessening cooperation. Thismight occur, for example, because providing strongerincentives for own goals makes people more focused onthese and less willing to devote time to doing things tohelp other people. This is shown in Figure 8. The frontierin the figure shows the maximum initiative available forany specific amount of cooperation, given a level of expen-diture on inducing these. Devoting more resources pre-sumably pushes the frontier out, allowing more of bothdesired behaviors.


107

In this framework, arm’s length market dealings typi-cally give the maximal incentives to pursue own goals andso the most initiative, but they provide little incentive forcooperation. So organizing through market arrangementsgives a point on (or close to) the vertical axis. The idealizedrational bureaucracy with centralized decision-making andweak performance incentives arguably would get littleinitiative but might achieve high levels of cooperation.Actual organizations get less cooperation than the ideal-ized hierarchy but more than the market. Of course, realcompanies need more cooperation than arm’s length deal-ings would induce, because interdependencies abound.The cost is that they do not generate the initiative thatmarket organization would generate.

A well-designed and well-managed organization will beon the frontier—otherwise, it could get more of bothcooperation and initiative for free. Of course, this does notmean that actual organizations are there—this is one ofthe big tasks of management! The design of the organiza-tion determines the mix of cooperation and initiative thatwill actually be achieved. The people, the architecture, the


108

Cooperation

Initi

ativ

eFrontier

Market

Idealizedhierarchy

Figure 8. Cooperation and initiative

processes and routines, and the culture all affect thebehavior that is induced. Where an organization shouldwant to position itself depends on the activities it is under-taking and what it is trying to accomplish, that is, on itsstrategy. Correspondingly, a shift in strategy might beexpected to lead to a change in organization. The strategywill also determine how much the firm should want tospend on organization, and thus where the frontier will lie.

For example, beginning in the late 1980s ABB AseaBrown Boveri attempted a complex strategy that requiredit to achieve, simultaneously, global efficiency, respon-siveness to the particulars of each of the hundreds of dis-tinct national product markets in which it competed, andthe spread of learning worldwide. All this had to be donewhile merging and integrating two predecessor compa-nies and numerous acquisitions. To do this, ABB neededboth strong initiative from its managers in pursuit of theirown objectives and, at the same time, a willingness to helpout other units and contribute to the whole.

This imperative led ABB to adopt a very complex andexpensive organizational form. ABB was made up of some1,300 operating companies, each with its own incomestatement and balance sheet. The operating companies inturn contained some 5,000 profit centers. The small size ofthe business units (the profit centers averaged perhapsthirty-five employees each) meant that responsibility couldbe clearly assigned and that the people in a unit couldreadily see the impact of their actions on results. Thiscontributed to generating initiative in pursuit of unit per-formance. A powerful financial and operational reportingsystem gave senior managers the information necessary totrack results and to motivate and control behavior.


109

Each operating company was matrixed under a dualreporting relationship. On the one hand, the manager ofan individual ABB company—for example, the steam tur-bine manufacturing operation in Germany—reported toa country head whose role was to ensure local responsive-ness and to coordinate across the ABB companies in thecountry, thus supporting cooperation on this dimension.Since many of the customers were governments orgovernment-owned electric and railroad companies, thislocal responsiveness was especially crucial. The countryheads also played a major role in postmerger integration.At the same time, the individual company head alsoreported to a global product area executive, who sought toachieve global efficiency by coordinating investments andallocating production across countries and to promotelearning by developing cross-national linkages amongengineers and functional specialists. Both bosses had a sayin performance evaluations and compensation.

Holding the whole together was a cadre of global man-agers who traveled almost constantly among differentbusiness units. Particularly important among these werethe Executive Committee members, each of whom over-saw a variety of product areas and countries. This designrequired huge amounts of managerial time and energy:CEO Percy Barnevik famously quipped that he was in theoffice two days a week—Saturday and Sunday. Such adesign would not have made sense under a less ambitiousstrategy.

As its strategy evolved, ABB adjusted its organization toaffect the mix of behaviors its people supplied. In the early1990s, the strategy moved to accentuate expansion inEastern Europe and Asia while correspondingly shrinking


110

the level of production in the West. This required morecooperation from the country heads in Western Europe,who would lose employees and investment and, withthem, influence inside the firm and with the national gov-ernments where they operated. The response was to cre-ate a regional structure at the top level of the firm, so thatone top manager oversaw all of the European countryheads and could make sure that the managers in WesternEurope supported the efforts in the East. This was toensure that the needed cooperation of the country man-agers could be induced.

A change in the environment can also change theneeded mix of cooperation and initiative. Johnson andJohnson, the huge pharmaceuticals, medical equipment,and consumer health care products company, organizeditself to generate very high levels of initiative (Pearson andHurstak 1992). It was divided internally into roughly150 separate, product-based companies, each fully self-sufficient and fiercely independent. In the 1980s, therewere thirteen different J&J companies serving the hospitalmarket for medical equipment, each aligned with a differ-ent medical or surgical department. These units had theirown sales forces, their own logistics and distribution, andtheir own billing. This was inconvenient and inefficientfor their customers, which under cost pressures had trans-ferred procurement from the individual departments tocentral purchasing managers. The obvious solution wasfor J&J to consolidate at least the distribution and billingfor the thirteen companies. Yet it took J&J fifteen years toeven try to establish such a system, because it wouldrequire more cooperation from the companies and threat-ened to undercut their independence and initiative.


111

Note that technological and organizational innovationscan move the cooperation–initiative frontier, making itpossible to have more cooperation and more initiativesimultaneously. Information technology is an obvioussource of such movements. By allowing finer perfor-mance measurement and better communications, it facil-itates getting more initiative (through lowering the cost ofproviding incentives) and more cooperation (by makingcoordination easier and increasing contact among units).Another important determinant is the technology oforganization and management. For example, the develop-ment of the multidivisional form for internal organizationand of long-term relations with suppliers are both inno-vations in management that shifted the frontier.

Using this framework to examine a company’s organi-zation design does not require actually measuring theamounts of cooperation and initiative that are being gen-erated. Rather, it simply requires evaluating the sorts ofbehavior being shown against what is needed. Two exam-ples illustrate this.

At BP Exploration, the disaggregation of the businessinto discrete, empowered business units with clear per-formance responsibilities and the elimination of layers ofmiddle managers and central functional staffs had theintended effect of increasing initiative, which JohnBrowne saw as critical to improving performance. But thechanges also created a great need for the business units tocooperate in sharing best practice and in supporting oneanother in solving technical and commercial problems—activities that were previously handled by the center, butthat it now lacked the resources to undertake. The solu-tion was to link collections of business units facing similar


112

technical and commercial challenges into peer groups,each of which had around ten members. The members ofthese peer groups developed norms of sharing informa-tion and of helping one another resolve problems. Centralto this was the practice of peer assists, under which a unitfacing technical or business difficulties could call on otherunits and they would respond by sending experts to helpresolve the problem. It became a strong norm to ask forhelp if in trouble and to respond wholeheartedly if askedfor help. This cooperation was crucial to BP’s success.

Novo Nordisk, one of the two global leaders in diabetes-treatment drugs, faced a crisis in 1992 when communica-tion problems inside the firm led it to be out of compliancewith new U.S. Food and Drug Administration regulations(Kamper, Podolny, and Roberts 2000). Novo had to with-draw temporarily from the U.S. market and to destroy alarge amount of insulin. In response to the crisis, it insti-tuted a new organizational design. It continued to allowlarge amounts of discretion to individual units in deter-mining the exact routines and procedures they would use,but it increased their accountability for adhering to broadcompany policies. The intent was to maintain the initiativethat had marked the highly decentralized organization butto get more cooperation in supporting overall policy andperformance. Crucial to this was the creation of a group of“facilitators.” The facilitators were experienced managersfrom around the organization who were charged with theinternal audit function of providing assurance that individ-ual units were adhering to company policies. When a unitwas found not to be in compliance, the facilitators wouldagree with the unit head on the needed changes. In thisprocess they came to know managers all over the company


113

and they became very knowledgeable about the experiencesdifferent managers had had. This allowed them to connectmanagers facing problems with others who had resolvedsimilar issues and to help the managers build networks thatwould support cooperative behavior.

Generally, the problems of effectively measuring coop-eration mean that it is difficult to provide formal incen-tives for it. Thus, to induce cooperation, it is better tolook to the softer elements of organization, includingsocial networks and norms, as was done at BPX and NovoNordisk. Meanwhile, architecture and process can bemanipulated in the first instance to encourage initiative.

Many firms have, in fact, decided that they are not get-ting enough initiative, that increased competitiondemands that their people work more effectively in deliv-ering performance in their jobs. Correspondingly, theyare moving towards more market-like solutions to theorganization problem. Within the organization theseinvolve delayering, using stronger performance pay,empowering managers, creating clearer business unitboundaries, and ceding authority over operations andeven elements of strategy to the units. These changesinvolve moving upwards in terms of Figure 8, gettingmore initiative, but they may also have the effect of mov-ing left, reducing the cooperation. To counter this effect,the firms are seeking new ways of relating to and moti-vating their employees and of connecting their people toone another. These firms are also moving activitiesacross their boundaries via outsourcing, spin-offs, andcarve-outs, which also typically increase initiative whileperhaps limiting the willingness to cooperate. At thesame time, they are not settling for the inadequate


114

cooperation that arm’s length market arrangementsusually engender. Instead, they are seeking more coop-eration by using novel ways to connect to suppliers, cus-tomers, and other firms, including relational contracts,joint ventures, and alliances. We will explore thesechanges more in later chapters.

Notes

1. This is not a claim about historical patterns—that marketswere tried but failed, and then firms were created. Rather,that the division of activity between firm and market haschanged over time in complex ways.

2. See Milgrom and Roberts (1992: chapters 5, 6, and 9) fora more detailed discussion of many of the topics addressedin this section.

3. This term originates in the insurance industry, where itrefers to the fact that those who know they are bad risksare more likely than the average person to buy insurance.

4. This term also comes from the insurance industry, whereit refers to the tendency of people with insurance to takegreater risks than they would if they bore the full impactof their choices.

5. For explorations of contracting with costly enforcement,see Doornik (2002, 2003).

6. There is a large literature in economics on incompletecontracts and renegotiation. See Williamson (1975, 1985),Grossman and Hart (1986), Hart (1995), Hart andHolmström (1987), and Hart and Moore (1990) for someof the early explorations of the institutional effects of thesefeatures. Later work is mathematically more demanding.Milgrom and Roberts (1992: 127–33) has a discussion


115

that, while still nontechnical, is more complete than thatoffered here.

7. Myerson and Satterthwaite (1983) show that, unless thereis common knowledge between buyer and seller that thebuyer’s willingness to pay for the object on sale exceeds itsvalue to the seller, there must be inefficiency. This canmanifest itself in trades that do not occur although theyshould, or in protracted negotiations and costly delays inreaching agreements.

8. See Milgrom and Roberts (1992: chapters 2 and 5) formore detail.

9. The strong form of the argument is that efficiency alonedetermines what will occur. This is true, however, only ifthe size of the pie is independent of its distribution:Formally, that there are no income effects and that wealthis freely transferable. These may be decent assumptions inmany circumstances, especially in thinking about relationsbetween firms, and theorizing built on them has been verysuccessful in explaining observed practice. When they fail,however, efficiency alone does not fully determine the pat-terns of organization and the allocations of resources thatare viable. The predictive power of the theory is thenreduced.

10. In contrast, “rents” are returns that are in excess of theminimum needed to attract a resource to a particular usein the first place. Note that once an investment is sunk,some of the returns may be quasirents even though therewere no rents (excess returns) being earned ex ante. Thissituation occurs exactly when there is specificity but com-petition obtains before the lock-in occurs.

11. For more on the hold-up problem, see Milgrom andRoberts (1992: 136–9).


116

12. See, for example, Monteverde and Teece (1982), Masten(1984), Joskow (1985, 1987, 1988), and Masten, Meehan,and Snyder (1989) for studies where asset specificity seemsto explain patterns of vertical integration and long-termcontracting.

13. Unless the measures are completely objective, which israre, and contractible, in the sense that third parties canobserve them, which is necessary for courts to enforcecontracts written upon them.

14. See Hart (1995) for an exposition of these ideas andWhinston (2003) for an evaluation of their empiricalimplications.

15. Concluding that the incentives inside the unified firm areactually weaker requires an additional assumption: Thatthe marginal impact of the investments is greater when therelationship is intact than if it breaks up. In particular, thismeans that the investments should not have more of aneffect at the margin on outside opportunities than on theperson’s value in the firm.

16. Formally, if the marginal cost of doing one increases in theamount of the other being undertaken.


117

118

4

Motivation in the Modern Firm

Firms, like other economic organizations, serve tocoordinate the actions of groups of people and to motivatethem to carry out needed activities. The problem of moti-vating people in organizations comes from the fact thattheir own self-interest may not automatically lead them toact in the ways that the organization would want. Thisdivergence of interests arises because the individualmembers of an organization typically do not bear all thecosts and benefits of the actions they take and the deci-sions they make within the organization. Consequently,when they make decisions—about how to spend theirtime, how hard to work and on what, what risks to take—the choices that appear best from their personal point ofview may not maximize the total value generated for theorganization. Even if they are cognizant of the largerinterests, they may not automatically take these fully intoaccount.

From an organization design perspective, the motiva-tion problem is to shape the organization—the people,the architecture, the routines and processes, and theculture—to bring a closer alignment of interests between

Motivation

119

the organization and its members and thereby increasethe efficiency of the choices they make. In this task, thedesigner in a firm can have access to all the levers of orga-nization design—people, architecture, processes, andculture—and in many contexts all these should be used.In particular, motivation is not just a matter of monetaryincentives, as important as these are in some cases.1

The Source and Nature of MotivationProblems

In general, motivation or incentive problems arise whenindividuals’2 organizational decisions and actions affectothers in ways that the individual does not fully take intoaccount—when there are externalities. There are basicallytwo ways that a disparity can arise between the costs andbenefits that an individual bears versus those that accrueto the organization as a whole. Perhaps most often, indi-viduals in organizations receive only a small fraction ofthe benefits that result from their taking various actions,but bear a disproportionate part of the costs that areinvolved. In this case, their decisions are likely to involvetoo little activity for organizational efficiency. The otherpossibility is that the fraction of the benefits that they getexceeds the share of costs that they bear. Then they arelikely to choose too much of the activity in question.

In the simplest example of the first sort of problem,someone employed for an hourly wage or a fixed salarydirectly experiences the physical and emotional effects ofworking harder and longer. The direct gains resultingfrom his extra effort are the increased output he generates,

Motivation

120

but these gains accrue to the firm, not the worker, whoseearnings do not change. There may, of course, be somegains to him: A decreased likelihood of being disciplinedor fired, a possible increase in the chance of promotion ora raise, and perhaps social approval (or opprobrium!) fromfellow workers. He might also experience some feelings ofindividual satisfaction, from a job better done. Yet,because he does not directly receive the full benefits of theincreased output, he is unlikely to be motivated to work ashard as he would have if he received the full fruits of hislabor. Indeed, there may be temptation to slack off, andif the shirking is not too obvious and egregious, he isunlikely to bear any great consequences. Thus, there is areal temptation not to work too hard—certainly not ashard as might maximize total value created.

Similarly, a manager may avoid taking risks that areworthwhile from the point of view of the shareholders.This happens because she receives little of the gains thatwould accrue to success, but faces the risk to her career thatfailure would bring. As another example, a shareholder whodoes a diligent job of monitoring corporate managementbears the all the costs of this activity. However, the benefitsin terms of improved performance are shared among allthose holding the firm’s stock. In this case, we wouldexpect less monitoring than would be optimal.

The motivation problem can also arise because thegains accrue to the actor but the costs are borne elsewherein the organization. For example, a CEO who enjoys theincreased status that comes from heading a bigger com-pany may be tempted to make acquisitions, even if theydestroy value.3 The executive gets the benefits, and thecosts are borne by shareholders. Note that other members

of the organization may support the CEO in this empirebuilding because their career prospects are enhanced bythe firm’s growth.

It is important to understand why motivation problemsarise in managed organizations—why there are systematicdivergences between the costs and benefits that accrue tothe organization and those that its decision-makers face.After all, we do not worry about motivation issues in alltransactions, even though the interests of the various par-ties are quite different and, at least regarding the price,are completely opposed. Financial market traders usuallydo not worry about having to motivate other tradersproperly when they trade yen for dollars in a foreignexchange transaction. Consumers do not worry aboutother parties’ motivation when they buy a can of brandedsoup from a grocery chain. Yet, when a firm seeks to buythe labor services from employees and contractors, itmust be very concerned about motivation.

The key difference is in the possibilities for using con-tracts or reputations to guide behavior in situations whereinterests differ and where the actions of one party have animpact on the well-being of others. If there is no diver-gence of interests, then there is no problem, because sim-ple self-interest will lead the parties to act in one another’scommon best interests. Even when interests diverge, if itis easy to devise and enforce contracts that induce thedesired behavior, then there is no real motivation problem.The contract, if well designed, will lead the parties to actin ways that maximize total value creation. Moreover, ifexplicit contracts cannot be written that adequately guidebehavior, reputation mechanisms can sometimes substi-tute for them and take care of motivation. But if interests

Motivation

121

differ, and if neither contracts nor reputation considerationsare fully effective, then there are significant motivationalissues.

In the foreign exchange example, it is clear to both sidesexactly what is required to meet the terms of the agree-ment to trade, and these conditions can be simply includedin a contract. Further, it is a relatively straightforward mat-ter to establish whether each side has adhered to the agree-ment. Thus, courts or other third parties (like arbitrationbodies) can enforce these agreements easily. Moreover, if atrader fails to abide by a contract, other traders on theexchange will know and can avoid dealing with the cheaterin the future. Commodity trade generally is relatively freefrom motivational problems, as are dealings where reputa-tions can effectively shape behavior. Pursuit of self-interestdoes not lead to any remediable inefficiency.

In the soup example, in contrast, there is no contractabout the recipe and processes used in making the soup,and until the customer gets it home and tries it, no oneknows if it will be to his taste. This might seem an oppor-tunity for the manufacturer to cheat on quality to savecost or to lie about the contents of the can to increasedemand. However, most of the actions the manufacturermight take that would have the biggest negative impact onthe customer are deterred by regulatory inspections andthe possibility of legal action. Meanwhile, the manufac-turer’s concern with repeat sales discourages its misrep-resenting quality in less dangerous ways. Regarding theretailer, there is little the store might do that could nega-tively affect the consumer and that would be worthwhile,so there is no reason to worry about its behavior. The oneexception might be that the store might charge very high

Motivation

122

prices, but reasonable levels of competition will take careof this too.

Things are different in managed organizations.Decisions and actions affect other parties, interests aretypically not fully aligned, contracting possibilities arelimited, and reputations are only partially effective.Indeed, these differences actually go a long way to explain-ing why firms exist!

A central, pervasive instance of such difficultiesinvolves limited observability of behavior and the conse-quent moral hazard. Individuals’ intellectual and physicalefforts at their jobs create benefits for the organization asa whole (in lower costs, better sales, reduced risk, andimproved reputation), but those efforts usually are notfreely observable and precisely measurable in a timelyfashion. Because the organization cannot tell exactly whatthe individual actually did, an enforceable contract can-not be written that would specify what is to be done andwhat rewards and punishments are to follow from adher-ing to or violating the contract. The limited observabilityof actions also undercuts the possibility for reputationconcerns to guide behavior. Since others cannot observethe individual’s actions, a reputation for behaving wellcannot be easily developed.4 This is the classic contextconsidered in the economic theory of agency. The exam-ples given of the worker deciding how hard to pushhimself or herself and the manager accepting or rejectingrisky projects are examples of this phenomenon, and infact the basic problem is pervasive.

In such contexts, any formal incentives have to be basedon noisy, imprecise indicators of what the agent has done.A prime example is using imprecise measures of behavior,

Motivation

123

such as intermittent monitoring, as a basis for performanceevaluation. Another is basing rewards on outcomes that areonly partially determined by the actions in question. As wewill discuss below, while such performance pay is increas-ingly popular, it has its problems.

A related motivational problem is known as free riding.It arises in situations where several individuals can con-tribute to some outcome but all share the benefits. Anexample is protection and promotion of a common brand.All the business units using the brand gain when any oneof them expends resources to defend or enhance it. Sinceno unit receives the full benefits that result from itsexpenditures, each is likely to under-invest. If the actionstaken with respect to the brand are hard to monitor, pre-venting the free riding may be difficult. As well, since thevalue any unit receives from the brand may be hard forothers to determine, it may be difficult to establish anappropriate amount for each to contribute to the brand.This opens the door for misrepresentation aimed atreducing any mandated contributions.

Another example of free riding comes in team-basedwork where the efforts of many individuals contribute tosome final result upon which rewards for the group arebased. It is tempting for each person in the team to do lessthan his share of the total needed for efficiency, becauseeach one incurs all the costs of any extra contributionsmade and yet gets only a fraction of the incremental bene-fits generated. Thus, unless the team members can easilymonitor one another’s contributions, each has an incentiveto stop contributing when the additional individual bene-fits match the additional costs of contributing more. But atthis point the total additional benefits accruing to everyone

Motivation

124

far exceed the costs of a single individual contributingmore. Thus, weighing the full benefits against the costswould call for much higher contributions. The resultinginefficiency means there is a motivation problem.

Limited observability of actions is not the only sourceof motivation problems. In particular, suppose that behav-ior is directly observed by all the affected parties, so thatthey all know what has in fact occurred, but that it is notpossible to establish the facts for an outside party. Thencontracts prescribing (or proscribing) particular actionscannot be enforced by outside third parties, and so theycannot function to motivate efficiently. Another problemarises when the parties fail to specify fully in their initialagreement what should be done in different circum-stances. This might occur because the parties did notforesee all the relevant possibilities, or because normallanguage is not sufficiently precise to distinguish differentcontingencies unambiguously, or because it was simplyjudged to be too costly to write such a detailed contract.Then contracts cannot fully specify the desired behaviorin every circumstance and so cannot fully resolve motiva-tion problems. Again, linking rewards to observable, veri-fiable outcomes (rather than behavior) may provide somedesirable incentives, but this can be even more problem-atic than in the earlier case. Reputations may be the onlyreally effective mechanism here.

A further difficulty arises when the actions that areactually taken are based on information that is availableonly to the person taking the actions. This situation is alsowidespread: Indeed, the power to make decisions is veryoften given precisely to those who are best informed andmost expert. A divergence of interests in such situations

Motivation

125

causes difficulties even if the actual actions themselvesare fully observable and verifiable by courts or other thirdparties. The problem is that it may be hard to judge if theactions were really in the best interests of the organizationor instead were chosen to advance the interests of theexpert. This is the case in the example of the empire-building CEO—there is almost always some story thatcan be told about how the latest acquisition is strategicand value enhancing, no matter how bad it actually is forshareholders. Once more, outcomes-based rewards (e.g.rewards tied to the stock price) are often a partial solu-tion, but they do not fully solve the problem.

All of these situations fall under the rubric of “agencyproblems.” There is now a very large literature on agencyin economics, and it has been applied widely in manage-ment studies. Most of this work deals with explicit incen-tive contacting, and we will review some of the highlightsof this work as a background for its application in laterchapters.

Simple Agency Theory5

The simplest agency model involves a single individual,called the agent, who acts on behalf of another, called theprincipal. (For clarity, we will treat the principal as female“she,” and the agent as male, “he.”) Examples include anemployee (agent) and employer (principal); a board mem-ber and the shareholders; or a contractor, lawyer, or brokerand the client. The returns to the agent’s actions accrue tothe principal except for those personal costs that the agentdirectly bears. For example, the employee’s output goes to

Motivation

126

the firm, which also absorbs the costs of materials andequipment, but the employee has to bear the costs of theeffort he takes. For simplicity we call the action being taken“effort provision,” but numerous other interpretations arepossible. What is crucial is that, other things being equal, atthe margin the agent prefers to provide less of the actiontaken on the principal’s behalf while the principal prefersmore. In particular, the agent is assumed to prefer workingless than would be efficient from the point of view of max-imizing the total benefits produced minus the costs of hisaction. The agent, in fact, may be willing to exert a non-trivial amount of effort without any explicit incentives, butproviding more effort than this imposes costs on him. Thiscreates a conflict of interest that underlies the motivationproblem in an agency relationship.

If the agent’s choice of effort were observable and ver-ifiable to the courts and if the desired action could bedetermined and described before the fact, then the twoparties could simply contract on the action to be taken.The principal would pay the agent for the effort he pro-vided for her benefit and, in light of the amount neededto compensate him for different levels of effort, she woulddecide the amount of effort to buy. (We refer to the effortchoice that would be made in this context as the “fullinformation” level. Since it would be value maximizing,we also call it the “first-best” level.) If the agent failed todeliver, then he would not be paid. Then he would, infact, find it worthwhile to act as the principal desired, andthere would be no significant motivation problem.

To have a motivation problem in this simplest, bare-bones model we assume then that the principal cannotdirectly observe the agent’s action. This is a reasonable

Motivation

127

assumption in many organizational contexts. We alsoinitially assume that some noisy signal of the action isobserved and can be contracted upon—otherwise, thereis no contractual way to provide any incentives. The mea-sure is assumed to vary with the agent’s effort, so observ-ing it provides information about the effort choice, but italso has an element of random variability that cannot beremoved. For example, the principal might attempt tomonitor behavior but cannot do so with perfect accuracy,observing instead the effort confounded with some ran-dom measurement error. This error might simply be thatthe monitoring is not continual. Then actual behavior isnot fully observed, and instead there is an episodic sam-pling of what was done. Alternatively, some outcome likeproduction volumes, costs, sales revenue, or profit mightbe observed, but this outcome might be determined bothby the agent’s action and other factors that are notobserved, such as the actions of other parties, or the ran-dom performance of a machine, or the state of demand.

The randomness of the performance measure meansthat the agent may supply relatively low effort but, by luck,the signal on which the reward is based may take on highvalues anyway. Similarly, the agent might work very hard,and yet the observed performance measure might be lowbecause of the effect of the randomness.6 Thus, rewardsbased on the imperfect measures will have an element ofrandomness that is beyond the agent’s (or the principal’s)control.

Of course, if the agent bore the full benefits of hisactions as well as the costs, he would take the efficientchoices. This might seem to suggest that rewards oughtto be structured to reflect the full impact of changing the

Motivation

128

effort choice on the benefits generated, which the agentwill then compare to the effect on the costs borne. Forexample, if the agent were the manager of a firm ownedby the principal, selling the firm to him would cause himto bear all the costs and benefits of his actions. Hewould then certainly face the right incentives. While thissolution is sometimes possible and attractive, there aretwo principal reasons why it might not be feasible ordesirable to have the agent bear the full benefits as well asthe costs of his actions.

The first reason is that the benefits might be uncertainwhile the agent is risk-averse, so that receiving an uncer-tain payment is less attractive to him than getting itsexpected value for sure. Then having him bear the full—but uncertain—impact of his actions is inefficient becauseit means that he shoulders all the risk arising from theuncertainty in the benefits, while the risk-absorbing abilityof the principal is not used and goes to waste. It would bebetter to share the variability of returns between the twoparties, because this lowers the total cost of the risk beingborne. Indeed, if the principal is risk-neutral, then sheshould ideally bear all the risk, because doing so is costlessfor her. The problem is that performance varies not justwith random luck but also with the agent’s choices in waysthat cannot be disentangled by the principal because theagent’s actions are not observed. Then having the principalabsorb some of the variability in returns means that someof the impact of the agent’s choice of actions necessarilymust also be borne by her. Thus, the agent does not face allthe costs and benefits involved in selecting his effort level.

The desirability of risk sharing remains an importantfactor even when the agent is not being asked to bear all the

Motivation

129

risk and when the noisy measure being used in contractingis something other than the returns that are generated.

The second reason that the agent may not be able tobear the full marginal returns of his effort choice is thathe is financially constrained. For example, perhaps hecannot cover any negative returns that might arise. In thiscase, selling the firm to the agent either is not possible or,if he has the option of bankruptcy, fails in fact to makehim bear the full marginal costs and benefits of hisactions. More generally, if there is a minimum payment(perhaps zero, or even negative) the agent must receive inany eventuality, any incentive scheme offered to the agentmust involve his always getting at least this amount, nomatter how bad the measured performance may be.Thus, it is possible that the agent cannot be made the fullresidual claimant.

In either case, once the agent cannot be the full residualclaimant, the principal’s problem is to design an incentivescheme that motivates the agent to provide effort inthe desired amount. The intensity of the incentives pro-vided—the extent to which rewards vary with the perfor-mance measure, and thus (given how the measure varieswith effort) how rewards vary with effort—determines theamount of effort the agent will choose to provide. Makingthe incentives more intense increases the expected returnto the agent from exerting more effort and so he will workharder. Assuming a risk-neutral principal, an optimalscheme simply induces the effort level that maximizes theprincipal’s expected returns net of what must be paid to theagent. Of course, the choice of the incentive scheme mustbe made in light of the fact that the level of effort actuallychosen will be what the scheme motivates the agent to

Motivation

130

provide—the principal cannot ignore the incentives thatthe reward scheme gives the agent. Further, the agent mustexpect to receive enough total compensation to be willingto work for the principal, rather than go elsewhere and pur-sue the next-best alternative use of his time and talents.

The theoretically ideal design involves a trade-off. Theexact nature of the trade-off depends on the reason theagent does not bear all the marginal costs and benefits.

In the limited liability case, providing more intenseincentives involves increasing the payment for goodresults without any offsetting decrease in payments whenthe performance appears bad (assuming the payment inbad states is already as low as is feasible). Thus, the costof getting more effort is that expected returns must bepassed to the agent, even though his pay is alreadyenough to attract him to the job. In fact, to get the first-best level of effort that the principal would buy from theagent under full observability might require giving theagent an expected payment that exceeds the expectedvalue of the gross returns. This effect may limit theintensity of incentives and result in a lower level of effortprovision than would obtain without an observabilityproblem.

In the more widely studied case on which we hence-forth focus, risk aversion is key. Giving more intenseincentives—for example, increasing the commission ratepaid to a salesperson—increases the effort providedbecause the returns to this effort increase. More intenseincentives also make the agent’s pay more risky, since agiven amount of random variation in the performancemeasure is now translated into a greater variance in pay. Ifthe principal is risk-neutral7 but the agent is risk-averse,

Motivation

131

however, having the agent bear any risk is costly. Evenwhen the principal is not risk-neutral, increasing theintensity of incentives eventually means shifting more riskonto the agent than would be desirable. The principal willhave to compensate the agent for the risk the latter is madeto bear, and so ultimately she bears these costs.

The principal’s problem in determining the desiredintensity of incentives then involves trading off the cost ofhaving to compensate the agent, both for exerting extraeffort and for bearing more risk, against the benefits gen-erated by the extra effort that is induced by strongerincentives. Generally, the solution is to have the agentface less intense incentives than are needed to induce thefull-observability, first-best level of effort, where the riskand the costs it would bring are irrelevant. At the sametime, the agent bears more risk than would be efficientabsent the need for incentives.

This model is very simple and stylized. Yet, it yields anumber of useful predictions and recommendationsabout the design of incentives. In particular, strongerincentives should be provided when the agent is less risk-averse and when the performance measures more accu-rately reflect what the agent actually did. The incentivesshould also be stronger the more valuable it is to the prin-cipal to induce higher levels of effort and the more easilythe agent can respond to strengthened incentives.8

The logic for each of these conclusions is a cost–benefitone. In two of the cases—the dependence of the inten-sity of incentives on the importance of inducing effortand on the responsiveness of the agent’s effort choice tostronger incentives—the logic is very simple. Regardingthe first, the greater the benefit of extra effort, the higher

Motivation

132

is the optimal amount of effort to induce, and thus themore intense should be the incentives. Similarly, the morethat effort responds to incentives, and thus the more extraoutput and value that result from strengthened incen-tives, the stronger should be the incentives.

These factors may account for the general pattern thatincentive intensity increases with the hierarchic level ofthe agent in most firms. Top executives’ decisions arguablyhave great impact on firm performance, and they have lotsof ways in which they can alter their behavior in responseto incentives. Meanwhile, less benefit typically flows fromincreasing the effort of a lower-level person, and suchemployees have fewer levers they can pull in responding toincentives. In the extreme, paying piece rates to a workeron an assembly line is pointless, because there is no way forthe individual worker to increase his output if others onthe line do not increase theirs.

The logic underlying the role of the degree of risk aver-sion is that the cost of the agent’s bearing extra risk is less-ened when he is less risk-averse. Since the incentive intensityis determined by trading off the costs of his bearing morerisk against the benefits from inducing more effort (net ofthe direct costs of the effort), the reduced marginal cost ofrisk-bearing leads to increasing the intensity of the incen-tives. To the extent that attitudes towards risk-taking dependon wealth, we might expect that those with high incomeswould be less risk-averse and better able to bear risks. Giventhat incomes and wealth tend to rise as one moves up thecorporate hierarchy, this then might be another reason whymore of the pay of top executives is at risk.

The logic for the role of the precision of performancemeasures is that, when the measures are more accurate

Motivation

133

and vary less in a purely random fashion, an increase inthe intensity of incentives brings less additional uncon-trolled risk in the agent’s reward, and thus less additionalrisk-bearing cost. It is then worthwhile to induce moreeffort via more intense incentives, because the cost ofdoing so is lower. Thus, in tasks where it is hard to mea-sure performance in an accurate and timely fashion, it isdesirable to use little in the way of explicit performanceincentives. In contrast, when the link between actions andobserved performance is clear and precise, very strongincentives can be given.

It is worth noting that there is actually a complemen-tarity between giving more intense incentives andimproving the performance measures. We have alreadyargued that better performance measures should lead tostronger incentives, but things go the other way too. Ifmore intense incentives are desired (e.g. because the valueof extra effort has increased), then it is more worthwhileto spend resources to improve the measures on which theincentive payments are based. This is because the value ofincreasing the precision comes from the reduced costs ofrisk bearing by the agent, and these costs are directlyrelated to the strength of the incentives being given. Sowith stronger incentives, the benefits of increasing theaccuracy with which performance is measured increase.

A further insight arises from this model. If explicitincentives are to be given at all, they probably should besubstantial: It is often better to give no explicit incentivesat all rather than weak ones. This is because there may bea discrete fixed cost of using performance pay, even ifthere are no administrative expenses in using a formalincentive scheme (which, of course, there likely are). If

Motivation

134

compensation is not dependent on measured results, thenthe net benefit of a given level of effort provision is simplythe net revenues of the firm, less the cost of that level ofeffort to the worker (for which the firm must compensatehim). If incentive pay is used to induce more effort, thenthe worker faces risk in his pay. This risk has a cost to theworker that is increasing in the personal cost of increas-ing the provision of effort.9 If the marginal cost of extraeffort is not zero, this term does not disappear as theamount of extra effort that the firm seeks to induce (and,correspondingly, the intensity of incentives) becomessmall. Thus, a little bit of incentive pay, although it mayinduce a little more effort, brings a discrete jump in costs.Only if the incentives are sufficiently strong, and enoughextra effort is thereby induced, will these fixed costs beovercome.

In particular, then, if the available measures are verypoor, so that only weak incentives might be offered in anycase, it may well be better to offer no explicit incentives atall. In this case, other means might be sought to motivate.One we will discuss below is “high commitment” humanresource management.

The Choice of Performance Measures

So far we have assumed there was a single measure onwhich pay could be based. Often, however, there are manypossible performance measures that are more or lessindicative of how much effort the agent has exerted. Forexample, both the stock price and accounting returns maycarry information about the quality of the job done by the

Motivation

135

senior managers of the firm, or it may be possible both tomonitor a worker’s behavior directly and to measure theresults he generates. An important issue is which of thesemultiple indicators should be used in designing perfor-mance contracts and how to use them.

The answer from agency theory is that pay shoulddepend on any freely available measure that is “informa-tive” about the agent’s effort provision (Holmström1979). Here “informative” means that taking properaccount of the measure allows a more precise inferenceabout the agent’s actual choice of effort than is possiblewithout it. Further, if a particular measure is not infor-mative in this sense, then it should not be used.

For example, the profits of an oil company depend notjust on how well its people have worked on finding oil,controlling costs, and generating sales but also, crucially,on the crude oil price. Rewarding employees on measuredprofits implies a lot of randomness in their pay becauseoil prices may double or triple from one year to the next.Thus, they could have done a miserable job and yet bewell rewarded just because oil prices rose. Equally, asuperb effort could go unrewarded because oil prices col-lapsed. This variability has no useful incentive propertiesand is costly. Instead, it would be better to remove theeffect of oil price changes in estimating performance.This was in fact done at BP plc, the global energy firm. Itemployed “self-help”—essentially the change in perfor-mance not attributable to changes in crude oil pricechanges (and exchange rates)—in performance pay. Thisamounts to using two measures, actual earnings and theoil price, to give a better indication of what the people inthe firm did.

Motivation

136

Note that this principle points to a need for subtlety inapplying the dictum that people should not be heldaccountable for factors they cannot control. Certainly ameasure that is not responsive to the agent’s effort shouldnot be used on its own as the only basis for rewards. Afterall, the point is to induce effort, and paying only on some-thing unrelated to what you are trying to induce accom-plishes nothing. But a variable that is correlated with thenoise in measured performance may be usefully employedin performance evaluation, even when it is not itselfdirectly tied to and reflective of effort. The reason is thatthe extra measure can be used to filter out some of theextraneous randomness. This is precisely what BP doesby taking out oil price effects.

Another application of this “InformativenessPrinciple” is to executive compensation. It is common forexecutives to receive bonuses that are based on account-ing earnings and also to have some of their pay tied to thestock price (either explicitly or through grants of stock oroptions). This makes no sense from the point of view ofagency theory unless the accounting numbers carry infor-mation that both is not reflected in the stock price and isrelevant for inferring how diligently the executives pur-sued their responsibilities. Under a common assumptionabout stock market price formation—that the pricesreflect all the publicly available information that is rele-vant for judging the value of the firm—publicly availableaccounting returns are unlikely to provide additionalinformation beyond the stock price about top executives’behavior. They, thus, should not figure into determiningtheir pay. The stock price should be a sufficient measure.On the other hand, accounting results may be quite

Motivation

137

informative about the efforts of functional or divisionalheads. For example, revenues might carry a lot of additionalinformation about the contributions of the marketing direc-tor. Thus, they should be used in his compensation, even ifhe is also rewarded with stock and options.

Using multiple measures to increase the overall preci-sion in measuring performance may seem complicated,but it need not be. A familiar example here is measuring asales organization by profitability—that is, revenues andcosts—rather than just revenues alone. Doing so, eventhough the sales people have no control over costs,encourages them to focus their efforts on more profitablesales, rather than on the biggest revenue generators.Rewarding the sales function on profitability is especiallyimportant if it controls pricing, because this rewardsystem encourages the sales force to account for costs inpricing decisions.

Another common example is compensating a salesperson both on his overall sales and on how he has donerelative to the average across the sales staff overall.Accounting for how others have done allows making abetter estimate of how hard the particular sales person hasactually worked by filtering out the impact of overallmarket conditions that affect every sales person’s results.Reducing the noise in measured performance reduces therandomness in the rewards, allowing stronger incentivesto be provided and more effort to be induced. Thus, arationale for comparative performance evaluationemerges. (Of course, this involves making the pay of oneagent depend on something not within the agent’s con-trol, namely the others’ performance levels.) Similarly,there is some evidence that CEOs’ compensation is

Motivation

138

responsive not just to their own firms’ performance butalso to how well they have done relative to comparablefirms in the same industry (Gibbons and Murphy 1990).This is consistent with the idea of using extra measures toreduce some of the variability in measured performance,and thus to permit more accurate assessments of behaviorand more intense incentives.

An extreme form of this comparative performance eval-uation occurs in “tournaments,” where rewards are basedsolely on the performance ranking of the different partici-pants, not on how well they performed in any absolutesense (Lazear and Rosen 1981). This system is familiarfrom golf and tennis contests and team sport leagues, but italso is used in sales contests, where the sales person with thehighest volume wins the prize. Promotions are also ofteneffectively a tournament: The best performer wins theprize of a higher salary and status and the opportunity tocompete for the next promotion. Tournaments can be veryuseful when it is hard to specify and measure results in aquantitative way on which pay can be based and yet it isstill clear who has done the best job. Gearing rewards toabsolute performance is then impossible, but the winnercan still be identified and rewarded. In some circum-stances, in fact, tournaments may be as effective motivatorsas explicit performance pay such as piece rates.

All these examples have involved performance measuresthat are more-or-less freely available because they are gen-erated in the normal course of business or for other pur-poses. If measures have to be developed and collected, thecosts of doing so are likely to involve a large fixed element.This may mean that only some of the potentially availablemeasures will be created and used. Another reason for not

Motivation

139

using all possible measures is that the resulting systemmay end up being too complicated for anyone to under-stand. In this case, it cannot motivate effectively. Thispossibility would seem to limit the potential effectivenessof schemes that call for large numbers of measures, suchas some variants of “Balanced Scorecard” systems.

When choosing among possible costly measures, firmsoften have to decide whether to measure inputs (behav-ior) or outputs (results). Which to choose depends onwhether the designers know in advance what the appro-priate behavior is (Prendergast 2000). If they know whatthe agent should do, then it is likely that input measureswill be effective: Specify what is supposed to be done,check to see if the agent has done it, pay him if he has, andfire him if he has not.10 There is no need for fancy incen-tive schemes, and this may explain why performance payis not more universal. On the other hand, if the designerdoes not know what ought to be done, then the designeralso does not know what behavior to specify and measure.In this case, it is better to measure results and leave itto the agent to figure out how to deliver these results.Note that, to the extent that environmental uncertaintyincreases the likelihood that the contract designer will notknow what behavior is desirable, then the presumptivenegative association between uncertainty and incentivesmay be reversed.

Multi-tasking in Agency Relationships

A richer set of insights arises when we extend the modelto recognize that the agent might spend his time on more

Motivation

140

than one activity that is useful to the principal. In thiscase, the principal needs both to motivate the overall pro-vision of effort and to shape its allocation among tasks.This becomes a significant problem in two different con-texts. One is when the desired activities compete for theagent’s time and attention, so that doing more of oneincreases the cost or difficulty of doing more of the other,while the available performance measures for the twoactivities are not of comparable accuracy or timeliness(Holmström and Milgrom 1991). In this case it becomesvery expensive to offer incentives that will induce theagent to devote significant levels of effort to both activi-ties. The second problem in the multi-tasking contextarises when there are not separate measures for the per-formance on the two tasks, but inducing the agent to do agood job on the one task requires very different incentivesthan are required for the other task (Athey and Roberts2001). When any available measures confound the resultsof the two activities in ways that do not permit disentan-gling, improving the incentives for one task may worsenthose for the other. In either of these contexts wheremulti-tasking is problematic, the solution to the motiva-tion problem often involves using other aspects of theorganization design, especially the design of jobs and theallocation of decision authority.

As an example of the first sort of multi-tasking prob-lem, the two tasks might be providing initiative and coop-eration, as discussed in the preceding chapter. Themeasure on initiative might be individual or business unitperformance, which might, in fact, be a reasonably accu-rate reflection of the effort and thought the agent pro-vided in this direction. Cooperation, however, is likely

Motivation

141

much harder to measure effectively, since the behavioritself may be hard to observe (especially if it involvesrefraining from activities that hurt other units) and theresults are going to be tied up in the performance of theother units and thus in the efforts their members haveexerted. Another example is delivering current perfor-mance and developing new business. Performance on thefirst is relatively easily measured, while information aboutthe quality of effort devoted to the latter is much less cer-tain and slower to emerge. The problem is to induce theagent to devote effort to both in the appropriate amounts.

The agent’s choice of how to spend his time—bothhow much effort to provide in total and how to divide itamong tasks—will be governed by the incentives that areprovided for the tasks. Suppose first that the activities donot compete for the agent’s time and attention—that thecosts to the agent of one activity are independent of thelevel of the other being undertaken. Further, supposethere are separate, independent (but imperfect) perfor-mance measures for each sort of effort. Then the choiceof the level of one activity to undertake will not affect thecost–reward trade-offs faced in choosing the level of theother. In this case, the incentives for each activity can beset independently to achieve whatever levels of the twosorts of effort are desired.

More often, however, the activities do compete at themargin, if only because time spent on one is not avail-able for the other. Then the agent’s working harder onone task raises the costs he experiences in providing moreof the second activity. In this case, increasing the rewardsto one activity will not only have the obvious direct effectof inducing the agent to devote more time and effort to that

Motivation

142

activity. It will also lead the agent to reduce his provisionof the other activity, because its marginal cost hasincreased while the returns to providing it have notchanged. This means the incentives for the two activitiesmust be designed in a coordinated way.

For example, suppose that all that matters to the agentis the total time and effort spent on the two tasks together,not how much goes into each task separately, and supposerewards are proportional to performance while expectedperformance in each task is proportional to the effortexerted on it. Then if the rewards for extra time spent oneach task are not identical, the optimal thing for the agentto do is to spend all his time on the better-rewarded task,because this maximizes his returns for any total expendi-ture of effort. To get attention paid to both tasks, thereturns to increasing effort on each must be equal.

The intensities of the incentives for different activitiesthus tend to be complements: Strengthening the incen-tives for one activity makes it more attractive tostrengthen the incentives for the others. Otherwise, theseother activities are ignored. Thus, the incentives offeredfor multiple tasks should all be intense together, or allrelatively muted.

More generally, even when the agent cares about howtime is divided among tasks (but doing more of one stillincreases the costs of doing more of the other), unless theincentives are appropriately balanced, the agent will tendto overemphasize the better-compensated activity andunder-supply the other. Indeed, cutting back on the badlypaid activity not only frees up time to spend on the betterpaid one, it reduces the cost to the agent of expandingsupply of the well-rewarded activity. In the extreme, the

Motivation

143

poorly compensated one will just be ignored, even thoughperformance on this dimension would, in fact, berewarded—just not enough.

Lincoln Electric provides an example of using strong,balanced incentives in a multi-tasking context. Lincoln isfamous for its extensive use of individual piece rates toreward production workers, paying them a fixed amountfor each item of output on which they complete theirassigned tasks. The piece rates provide very strong, directincentives to produce lots of output. The danger is that,when so strongly motivated to increase quantity, theworkers will stint on quality and will be unwilling to helpout in various ways that are crucial to the firm but takethem away from earning their piece rates. To prevent this,the workers are paid a bonus based on the quality of theirwork and other factors like the ideas they generate andhow cooperative they are. On average, the bonus doublesthe employees’ already substantial income from the piecerates. Moreover, the bonuses vary dramatically betweenworkers, depending on their assessed performance. Thesestrong, balanced incentives have helped Lincoln achieveunmatched productivity and a reputation for top qualitythat have led to an outstanding record of decades ofbusiness success.

Providing comparably intense incentives for differentactivities becomes problematic, however, when the availablemeasures of the two tasks differ greatly in their precision ortimeliness.

Generally, the more accurately performance in an activ-ity is measured, the less costly it is to provide strongerincentives for the activity in isolation and so to inducehigher effort on it. Good measures tend to lead to strong

Motivation

144

incentives, and poor measures lead to weak ones, becausestronger incentives lead to greater increases in risk-bearingcosts when the measures are worse.

Suppose then that the quality of the measures on twoactivities differs significantly, as in the example of control-ling costs in current operations and trying to develop ideasfor new business opportunities. Realized costs are proba-bly a pretty good measure of cost-control efforts, but theagent might easily work very hard at developing ideas andyet have little to show for it. Or the agent might have comeup with some innovative ideas, but assessing their valuewill take time. In this case, it may be extremely difficult orexpensive (in terms of the risk borne, for which the prin-cipal must compensate the agent) to give strong incentivesfor idea generation, even though it is cheap and easy toprovide intense incentives for cost control.

We have already seen that this is not be a problem if theactivities do not compete at the margin for the agent’stime and attention—if doing more of one task does notaffect how hard it is for the agent to increase the other.Then, the incentives for each can be set independently.The well-measured one will carry strong incentives andthe other task will offer weak ones. The strong incentiveswill induce lots of effort aimed at the well-rewarded task,and the low returns to effort in the other task mean it willnot get very much attention. But still each will be pro-vided at the level the principal sees as appropriate in lightof the costs and benefits.

If the tasks do compete at the margin, however, then giv-ing strong incentives for the well-measured one and weakincentives for the other will result in very little or no atten-tion being paid to the latter task, no matter how important

Motivation

145

it may be to the principal. (Note that the impossibility ofobserving the agent’s choices means the principal’s simplydirecting—or imploring—the agent to do both is not veryhelpful. The agent is still going to be inclined to make thechoices that are good for himself in light of the incentivesprovided.)

Giving intense incentives for some desirable activitiescan then be a very bad idea, because these become nega-tive incentives for other activities that cannot be similarlyrewarded. The general idea that you get what you measureand pay for has a very precise meaning here.

An example involves merit pay for teachers (Milgromand Roberts 1992: 230–1). A popular proposal would payU.S. public school teachers more if their students do bet-ter on standardized tests. In contrast, teachers’ pay now isusually based on credentials and experience, so the explicit,financial incentives for performance are quite weak(although intrinsic motivation is obviously real andimportant). Proponents of the proposed reform arguethat providing stronger incentives would lead to betterperformance by teachers and their pupils. In all likelihoodit would, in fact, lead teachers to do more of whatever ittakes to help their students do well on the tests, particu-larly if the performance element of pay were substantial.However, it would also likely lead them to spend muchless time and effort on things that are not measured onthe tests. Indeed, in California, where schools’ funding istied to student performance on standardized tests ofmathematics and reading, there are claims that teachershave de-emphasized teaching other subjects, even thoughtheir pay is not directly affected by the test results.11

Some of these other things may be very important. They

Motivation

146

include not only other academic subjects (which couldpossibly be included in the testing), but also things thatare hard to measure, like helping develop students’ char-acters, teaching ethical behavior, and encouraging goodcitizenship. Measuring what teachers do on these dimen-sions in a relatively precise and timely fashion seems veryproblematic. So merit pay based on test performance islikely to drive these out, although they are provided in theabsence of explicit incentives. Even worse, it might leadthe least scrupulous teachers to find inappropriate waysto ensure their students succeed, such as getting hold ofthe test questions in advance. There actually have beensome instances of such behavior in New York State, whereperformance on the state examinations at the end of highschool is hugely important.

Thus, if multi-tasking is desired, it may be best to sup-ply relatively weak incentives for both activities. Anyeffective multi-tasking incentive scheme must be bal-anced, offering similar rewards at the margin for eachtask, and if some tasks are poorly measured, then makingall the incentives relatively strong will entail unacceptablelevels of risk for the agent and correspondingly highcompensation costs for the principal. The weakness of theincentives then means they will not motivate the agentto exert huge amounts of effort on either task, even thewell-measured one to which he could be induced to devotesignificant effort if it were in isolation. Consequently, theperformance on this activity is likely to be much worsethan it would be in isolation. Yet, this may be better thangetting none of the other activity.12

Thus, it may be better to leave teachers’ pay largelyunrelated to students’ measured performance on tests.

Motivation

147

Similarly, it is likely a good idea to avoid piece rates inmanufacturing when contributions to quality cannot bemeasured in an accurate, timely fashion. And payingmanagers for current results can lead them to mortgagethe future of the business, especially if they expect tomove on to new jobs before the effects of their actionsbecome clear.

The suggestion that performance pay should beavoided for tasks where it might be easily and effectivelyused certainly runs counter to the message business man-agers have received in the last decades. Yet, the frustra-tions so many executives have expressed with the failureto get their people to generate new ideas and growthopportunities indicates that the strong incentives theyhave created for current performance may sometimes becounterproductive.

An obvious solution here might be to separate the twotasks between two different agents. With no multi-taskingproblem, one agent can be given intense incentives andthe other more muted ones. Sometimes this approach isfeasible, and it may even be the best solution, when theactivities call for very different talents or skills or draw ondifferent knowledge.

In other cases dividing responsibilities is very costly.Not only does a second agent need to be paid, but anysynergies between the tasks may be lost. For example,sales representatives may have opportunities to learnabout customer needs, and thus about new opportunitiesfor product development. It would then be useful to askthem both to sell the current products and to bringsuch ideas back to the firm. But this creates a multi-tasking problem because the possibilities for measuring

Motivation

148

performance on the two activities differ so much. A casestudy of the electronics parts business (Anderson 1985;Anderson and Schmittlein 1984) illustrates the solution.When bringing ideas back is crucial, firms use employees assales agents, paying them salaries (with little or no bonusfor extra sales) and asking them to pay attention both tocurrent sales and to the communication of ideas. This is ascheme with balanced, weak incentives for multi-tasking.When there are fewer opportunities for bringing backideas, outside sales representatives are used. They are givenstrong incentives for extra sales (as they must be in order toget them to focus on this firm’s products rather than thoseof other firms they represent) and there is no seriousattempt to induce them to develop and share informationabout customers.

In the extreme, separating the tasks is simply impossible:There is no way to make one worker responsible for thevolume produced while holding another responsiblefor the quality of the output the first one produces.When multi-tasking is unavoidable, it is very valuable to beable to increase the precision of the measures of the agent’sperformance on the poorly measured task. LincolnElectric’s effort to identify the individuals responsible forthe quality of each machine by having them stencil theirnames on the pieces they produce is illustrative of thisimperative.

The second sort of multi-tasking problem involvesthere not being adequate measures for the different tasksindividually. For example, the agents might be themanagers of two business units. Each manager needs toexert effort at leading the business unit, and each can alsomake decisions (about the brand, dealings with shared

Motivation

149

customers and suppliers, human resource policies, and soon) that may affect the returns to both units. Meanwhile,any available measures confound the results of the effortprovision and the decisions. For example, none of theusual accounting measures would permit separating outthe specific effects of the manager’s effort, of the deci-sions he took, and of the decisions that the other managermade that affected the first division’s costs and sales.Then it is not possible to provide incentives separately foreffort and decision-making—any reward scheme willsimultaneously affect the incentives for effort provisionand for choices. The difficulty is that the sort of incentivesthat would be used to induce lots of effort may inducepoor decision-making, while those that lead to good deci-sions do not motivate effort provision effectively. Thus,there are trade-offs.

To induce good decision-making on matters that havespillover effects on the other division, the manager’sincentives should reflect the performance of both units.This might be realized, for example, by a bonus basedon the aggregate profitability of the two. In this way,he will be induced to pay attention to the full impactof his choices. He will also have incentives to workhard to develop the right sort of investment opportun-ities and the information needed to make good choices.Of course, if his pay is based on overall profitability,the other unit’s performance enters positively into hisrewards.

On the other hand, to induce effort efficiently, eachmanager’s incentives should be based on narrow measuresof his division’s performance alone. In particular, theother division’s performance should not carry positive

Motivation

150

weight in determining the manager’s compensation.Indeed, if the objective were simply to induce effort, theother division’s measured performance might often beexpected to have a negative impact on the manager’s pay.If the performance of the two divisions is positively corre-lated because of the impact of common factors like generalbusiness conditions, then the Informativeness Principlemeans that comparative performance evaluation shouldbe used. This means that a manager’s performance isviewed less positively when the other division does better,and so the pay actually is negatively related to the otherdivision’s returns.

Thus, the incentives that induce good decision-makingare bad at inducing effort (because they put too muchextraneous risk on the manager). On the other hand,incentives that induce effort effectively lead the managerto ignore the impact of the decisions on other divisionsand even to prefer choices that impose costs on othergroups.

If the individual managers on their own must make thedecisions, then the design of the system simply balancesthe importance of high effort and good decisions.Getting better decisions means making managers be con-cerned with the other unit’s performance as well as theirown. But this implies loading the managers with avoid-able risk (from the other division’s performance) and theincreased costs of the risk lead to giving less intenseincentives. In turn, less effort is induced. The optimalscheme balances these gains and losses, with the point ofbalance determined by the relative importance of effortand decisions. If effort is more important, then incen-tives will be based primarily on the manager’s unit’s

Motivation

151

performance, and he will make decisions that are goodfor his unit but not necessarily good for the firm overall.If his decisions have a large impact, both on his unit andon the other, then his pay will be mostly on overall per-formance, and he will not be as highly motivated to workhard for his unit’s success.

If it is possible for others to make the decisions for aunit, then new options arise to design the decision-making process as well as the incentive schemes to getbetter performance on both dimensions. For example, thedesign might specify that a decision about a project aris-ing in one unit that affects another would be imple-mented if and only if both units agree to it. As a firstapproximation, whether a manager will agree depends onwhat the adoption will do to his pay, which in turn candepend either just on the manager’s own unit’s measuredperformance or on both units’ results. Since under anywell-designed incentives, each manager’s pay increases inthe manager’s own unit’s performance, any project that isaccepted under these rules will improve the measuredperformance of both units and presumably that of thefirm overall. However, units will tend to reject projectsthat hurt them, even if they increase overall value.Further, if comparative performance evaluation is used,then the fraction of projects accepted will be small. Thisis because projects that help one unit’s performance arebad from the other’s point of view. Thus, projects will beaccepted only if they give comparably positive returns toeach unit. As a consequence, this system is problematic ifprojects tend to have their main impact on the division inwhich they emerge (so that, under comparative perfor-mance evaluation, the other division will not like them)

Motivation

152

but do have direct effects on other units (so that they getto veto projects).

In this context, a better design would be that if bothunits agree, then the project is implemented, but if theydisagree, then the decision is referred to a third party.Other things being equal, this party should be givenincentives that ensure good decisions, while the unit man-agers get incentives that motivate them to work hard.They will then accept projects that improve the measuredperformance of both units (and thus of the firm as awhole, if these are the only affected units). Meanwhile,projects that are worthwhile overall but harmful to oneunit will be referred to the third party, who will acceptthem. This referral process is presumably costly, which iswhy it is worthwhile to let the affected units decide ifthey agree.

This latter process is in fact quite similar to the “con-currence” system that was used at IBM (Vance, Bhambri,and Wilson 1980). Under this arrangement, businessunits had to obtain the sign-off of any other affected unitsbefore they could implement projects. If concurrence waswithheld, then the implementation decision would bepassed to the units’ direct hierarchic superiors. If theycould not agree, it would pass further up until a commonsuperior with authority over both units was reached andmade the decision. This process could go as far as theexecutive committee, and it occasionally did.

Other designs encompassing incentive contracts anddecision processes can be optimal in other contexts. Thekey is that multi-tasking causes problems and that thesolution may involve multiple aspects of organizationdesign.

Motivation

153

Group Performance Pay

Many management experts have argued against individualperformance pay on a variety of grounds. A major one isthat it allegedly destroys cooperation and teamwork. Thiseffect may indeed be real, as the logic of multi-taskingmakes clear, and it is especially likely if performance isjudged on a relative basis. In fact, it may be easier to sub-vert a colleague’s performance than to improve one’s own.

Of course, to the extent that the group can be thoughtof as acting in concert, the preceding theory can beapplied to motivating a group using rewards based ontheir collective performance. A very direct example is thepractice in California of paying teams of tree-fruit pickersby the total number of boxes they fill. The problem is thatthere may be free riding, because the results of extraeffort by any group member are shared across the wholegroup. The resultant shirking may be effectively limited,however, by mutual monitoring within the group, espe-cially if the group is not too large. This is an argument forkeeping groups small for performance measurement andreward purposes. It is also necessary that a group norm ofworking hard be established and enforced by the groupon its members. Of course, a norm of group-wide shirk-ing might emerge instead. For example, Roy’s classicstudy of a machine shop (Roy 1952) documented howworkers physically punished colleagues who exerted toomuch effort and threatened to undermine the norm offeatherbedding. Managing this element of culture is thena crucial task for management.

These free-rider issues are especially relevant to thecommon practice of paying rewards based on overall

Motivation

154

corporate performance, whether through bonuses, profitsharing, stock grants, or option awards. In a typical largefirm with tens of thousands of employees, the free-riderproblem would seem especially intense. Suppose anemployee’s efforts would create an additional million dol-lars in earnings. Then this might lead, generously, to anextra hundred thousand into the bonus pool. At most afew thousand of this is likely to trickle down to theemployee responsible. More likely, he would get muchless than this. Thus, from a purely financial point of view,the employee will rationally undertake the extra effortonly if the costs are less than the extra bonus he can expect.Yet, the value of the extra effort to the organization ismany, many times this amount—a million dollars. Effortis rationally under-provided because the benefits areshared but the employee bears all the costs. The situationis even worse with stock-based pay.13 The employee’sshare of the extra value created by his actions is just theemployee’s fraction of the ownership of the outstandingstock. This is almost surely a trivially small number.

Stock-based pay thus seems remarkably inefficient as adirect motivation tool. Yet it is common, and not just atsenior executive levels. While there are certainly otherreasons for giving ownership claims to employees, such asencouraging them to invest in firm specific human capital(Roberts and Van den Steen 2001), it just does not seema sensible way to motivate effort.

A plausible explanation for its prevalence lies in its sup-porting norms of hard work and mutual monitoring. Ifstock ownership somehow changes employees’ mindsets,making them “think like owners,” then it could be an effec-tive motivational tool. This is especially likely if the firm

Motivation

155

combines this with a series of other measures in what isoften called a “high commitment” work system (see below).

Manipulation of Performance Measures

A second complaint against performance pay is that, toooften, the available measures are manipulable—the agentcan find ways to increase measured performance that areeasier than doing what is wanted. Then higher measuredperformance may be achieved by lowering the actual valuecreated for the organization. The examples here are all toonumerous. Paul Oyer (1998) has documented how payingbonuses for meeting annual performance targets leadsmanagers to accelerate or delay sales in the last quarter ofthe year in order to make the numbers while not “wasting”sales that exceed the target. This increases the agents’ pay,but cannot help the firm. At H. J. Heinz & Co, divisionmanagers were required to post steadily increasing resultsevery year in order to earn bonuses. They manipulated theaccounts to achieve their targets (Horngren 1999: 937–8).As the capital markets have come to demand steady earn-ings growth, such practices have become appallingly wide-spread. In two of the more egregious recent examples,industrial giants Enron and Worldcom each collapsed intobankruptcy after the exposure of the manipulations thatthey employed to hide their true performance and hold uptheir stock prices. Enron’s trickery was sophisticated, mis-representing the true balance sheet situation and profit-ability via asset transfers to nominally unrelated entitiesthat in fact were controlled by Enron executives. Restatingthe firm’s earnings to undo the chicanery reduced them by

Motivation

156

two-thirds from their reported levels. Worldcom’sapproach was simpler. It inflated its earnings by blatantlyrecording billions of dollars worth of current costs as cap-ital expenditures and by treating non-recurring receipts asif they were revenue that could be expected to be ongoing.These deceits were easier ways for the executives to bolsterthe value of their stock holdings than actually generatingthe returns honestly.

These examples involve misleading corporate superiorsand/or shareholders, but customers can be the targetvictims as well. Sears Roebuck in 1992 sought to motivatethe mechanics in its auto repair business by setting targetsfor the amount of work they did. The mechanicsresponded by telling customers that they needed steeringand suspension repairs that were in fact unnecessary.Customers could not easily verify the need for the repairsthemselves, and many paid for the unnecessary work.When the fraud was uncovered, Sears not only paid largefines, it lost much of the precious trust it had onceenjoyed among its customers.

In some of these instances, the problem is that the incen-tives are poorly designed. Giving a fixed bonus if and onlyif a target is met invites the agent to meet the target and nomore. This then can lead to the problems of manipulatingthe timing of sales and of doctoring internal accounts.

In other examples, monitoring some elements of behav-ior directly would have helped control the manipulation.Sears’ incentive scheme led to a major shift in the sort ofwork being done, with the extra business being concen-trated in repairs the need for which it is easy to misrepre-sent. If Sears’ managers had monitored the mix, theycould have caught the cheating. The top executives at

Motivation

157

Enron and Worldcom claimed ignorance of the accountingmanipulations. If they were indeed ignorant, theycertainly had not done a very diligent job of monitoring.

Making pay less responsive to the manipulable mea-sure is also an appropriate response (Baker 2000): If thepeople at Enron had not had such huge amounts tied upin the company’s stock and options, perhaps they wouldnot have been so eager to push up the stock price.

Another important way to limit the incentives formanipulation is to avoid formulaic reward schemes,replacing them with more subjective evaluations and bas-ing rewards on these. In fact, subjective evaluations andsubjectively determined pay represent an important alter-native to direct performance pay in providing incentives.

Subjective Evaluations

The basic theory of agency assumes that enforceablecontracts can be written on the observed performancemeasures, even though the actual desired behavior is notdirectly contractible. In many instances, however, theremay be information about performance that is availablebut not easily used in explicit contracts because it is toocomplex and too difficult to describe, or too hard to verify by third parties who might enforce the contract.Most subjective evaluations are of this form. Yet, suchinformation obviously might be useful for motivation ifrewards—whether pay, or promotions, or something lesstangible but still valued—can be tied to it.

The problem comes with the firm’s incentives actuallyto carry through on its promises to pay the appropriate

Motivation

158

rewards when they have been earned. There are at leasttwo sorts of difficulties.

First, once the job has been done, the principal maygain by reneging on the promised rewards. The mostobvious problem involves refusing to pay, claimingfalsely that performance was not satisfactory. The lack ofan explicit contract means that the cheated agent cannotgo to court to enforce the agreement. But if the princi-pal’s promises are not credible to the agent (as they cer-tainly will not be once the first reneging has occurred),they will not induce effort in the first place. Indeed, evenwithout the principal having yet cheated, if agentsunderstand her incentives they will be unlikely to trusther to pay them honestly and so will not work hard.

In a related vein, perceived arbitrariness and ambiguityin evaluations can undermine incentives. If the basis forthe subjective evaluation is unclear to the agent, then hecannot respond by altering his behavior in the desiredways. Bias and favoritism too can be a problem. If employ-ees believe that the rewards will flow to the favored, not tothe deserving, the promised rewards will not induce thedesired behavior.

It is also possible that the principal may be too for-giving, paying rewards that were not really earned. Thismay seem unlikely when the principal is acting on herown account, but it is certainly an issue when the prin-cipal is also an agent and is not well motivated. Forexample, the compensation committee of a board ofdirectors may reward the CEO handsomely despite thefirm’s miserable performance because they feel moreallegiance to the CEO than to the shareholders. Theyget away with this because free riding and informational

Motivation

159

asymmetries mean the shareholders do not and cannotmonitor the Board effectively.

The second major difficulty with subjective perfor-mance evaluation is that it may be subject to influenceactivities. Without a clearly articulated, mutually under-stood, unambiguous basis for evaluating performance,the agent has every reason to try to affect the principal’sjudgments and decisions. The agent can marshal argu-ments and evidence that, in fact, he did a great job, andthere is some reasonable expectation that this may swaydecisions. Moreover, if the evaluation involves any com-parative aspects, then the agent also has incentives totry to make others look bad. The danger is that the wholesystem becomes politicized, and the rewards are ultimatelygiven for success in special pleading, self-promotion, andsabotage, and not for doing one’s job. The perversity ofsuch incentives is clear.

The solution to many of these difficulties lies in theprincipal’s reputation (Baker, Gibbons, and Murphy1994). If the principal has a history of not acting oppor-tunistically, keeping its non-contractual promises, andnot being swayed by influence activities, then agents mayextrapolate this past behavior into the future. They maythen be motivated by the principal’s promises. Theyexpect her to keep her promises, because not doing sowould hurt her reputation and thus her ability to induceothers to behave as she wants in the future. In effect, herreputation becomes an asset that generates value byaffecting the behavior of both the principal and theagent.

Motivation

160

Reputations

Reputations can arise in many contexts and the samegeneral principles apply across them all. In particular,managers (agents) may be motivated to work hard and gen-erate good performance by their concern with their repu-tations in the labor market (Holmström 1982a; Gibbonsand Murphy 1992). Even when there is no explicit pay forperformance, doing well increases their market value andthis may affect their pay in the future. However, we willdevelop the analysis by continuing with the principal–agent framework and the principal’s use of her reputationin supporting motivation via non-contractual rewards.

If the reputation mechanism is to be effective, then ineach interaction the returns from maintaining an unsul-lied reputation must exceed the gain from violating trustand reneging on promises. The value of an intact reputa-tion comes from its leading others to trust the principal’spromises and so be willing to act as she would want—working diligently and cleverly in the current context.This means there must be future opportunities for her touse her reputation, that the future gains from using anintact reputation to affect behavior are significant relativeto the immediate gain from reneging, and that the princi-pal does not discount these future returns too heavily.

Frequently repeated dealings, either with the currentagent or with others, give the basis for future use of thereputation. This means that each time the parties interactthe principal must anticipate additional opportunities togain from the reputation. Consequently, there has to be at

Motivation

161

least some probability of the stream of dealings extendingindefinitely, and short time horizons are a major problemfor reputation formation and use. Meanwhile, increasingthe number and frequency of interactions helps supportgood behavior by increasing the returns to maintainingthe reputation.

For there to be significant future gains from the repu-tation, the agents who will deal with the principal in thefuture also must be able to observe and recognize cheat-ing whenever it occurs, even if they are not directlyaffected. Then they must be willing and able to punishthe cheating by behaving differently than they would haveif the promises had been kept. All this puts limits on thepossibility of using reputation. When performance ishard to describe, the measurements are subjective, andany promises are necessarily vague, knowing whether theimplicit contract has been breached is problematic. Didmy performance merit better treatment and more reward?Did yours? And will agents be willing to punish the prin-cipal when she apparently has cheated? Punishmentcannot be too costly for the punisher, or it will not beinvoked voluntarily. Here a sense of fair play, a desireto see justice done, or a psychological need for takingretribution can be helpful, because they lower the costs ofpunishing transgressors. They thus deter cheating.

Even large gains to inducing good behavior in thefuture by maintaining a reputation for fair dealing andkeeping promises will not be enough to motivate theprincipal if she discounts future returns very heavily.Even the most honorable principal, if faced with immi-nent bankruptcy, is likely to find that the immediate gainsfrom reneging on promised but non-contractual rewards

Motivation

162

are just too tempting. The principal recognizes that herreputation may be shot and that this will mean that anyfuture dealings may be more difficult. Still, if the optionsare keeping agreements and going broke or reneging andsurviving, the choice is clear.

While it seems clear that reputations in fact play a hugerole in reward systems within firms, some of the moststriking evidence on their use comes from trade ineconomies where the institutions of contract law are notwell established. John McMillan and ChristopherWoodruff (1999a, b) studied emerging private business inVietnam. Businesses there cannot count on courts toenforce contracts. Consequently, they rely on reputationsheavily. People in the same business meet regularly to dis-cuss which customers have refused to pay them, so that allmay avoid future dealings with the cheats. This raises thecost of cheating in a particular deal, because all sellerspunish the cheater, not just one. It thus supports betterbehavior. As well, large amounts of effort go into screen-ing potential customers before taking the risk of supply-ing them. Once a successful relationship has beenestablished between trading partners, the asset will beused repeatedly, even to the extent of their increasing thescope of business activities so as to be able to trade morewith one another. This leverages the scarce asset, trust.Moreover, it also supports more cooperative behavior ineach transaction, because more is at stake across the manydealings.

The value of reputation and the importance of thereputation-bearer’s being long-lived (so that it always hasmore opportunities to use the asset) suggest an advantageof organizing a permanent firm rather than leaving

Motivation

163

transactions to be market dealings between individuals(Kreps 1990). A corporation is, in principle, immortal.Further, it will have many more opportunities than anyindividual to use any reputation it develops. Thus, it maybe an effective mechanism for supporting efficient dealings.

PARC and Motivation

Formal agency theory has tended to emphasize contrac-tual responses to motivation. Yet, these are only some ofthe means that are available for motivating. In fact, all theaspects of organization can be employed, alone or in con-cert, as can managerial vision and strategy. We will illus-trate with a number of examples.

Managerial vision is a clear conception of a desirablefuture state of the world and the business. For example,Nokia’s vision in the early 1990s was that “Voice will gowireless.” Mobile phones would become ubiquitous,making it a very attractive business for Nokia. Steve Jobsand Steve Wozniak of Apple had a vision that the personalcomputer would be similarly ubiquitous. Having a leaderwhose vision is very clear and certain can be highly moti-vating for employees, inducing more effort and guidingits allocation, because they are sure of what will berewarded (Rotemberg and Saloner 2000; Van den Steen2002). Strategy can serve a similar function (Rotembergand Saloner 1994). If there is clarity about what the firmwill do and what it will not, well-placed effort is morelikely to be rewarded and, thus, to be provided.

Turning to the organization, first the people dimensioncan be used. Whom the firm attracts and selects as

Motivation

164

employees can have a tremendous effect on their motivation.Clearly, it should seek to attract people who are interested orchallenged by the work being done. This is intuitivelyclear—if people like their work, there is less of a problemmotivating them. More formally, this sort of matchingreduces the divergence of interests that underlies the moti-vation problem. Also, it is clearly advantageous to matchpeople with the rewards that can be offered. For example, ifit is desirable to provide intense incentives but performancemeasures are not very precise or results are hard to forecast,then it is important that the people put in these jobs notbe too risk-averse. If only non-monetary incentives canbe offered, then get people who value the rewards that arepossible.

In this context, it is important to recognize that theformal rewards systems offered will cause potentialemployees to self-select. Thus, an element of the organi-zational routine affects the people dimension of theorganization. For example, Safelite Auto Glass, which isin the business of replacing broken car windshields in thefield, instituted a piece rate to replace the hourly wages ithad previously paid installers. Productivity increased by44 percent within the year following the change. Half thegain was attributable to the motivation effect of the piecerate’s leading installers to work harder and faster. Theremaining 22 percent increase was due to selectioneffects. People who were willing to work hard were dif-ferentially attracted to Safelite, where their efforts wouldbe rewarded, and turnover among the most productiveworkers fell dramatically (Lazear 2000).

The importance of having a good fit between thepeople and the other elements of the organization became

Motivation

165

evident when Lincoln Electric made a number of foreignacquisitions in the early 1990s (Bartlett and O’Connell1998). Lincoln Electric’s pay system offers exceptionalmonetary rewards for production workers who are willingand able to work very hard. In its home base of Cleveland,the bargain is well understood, and self-selection is effec-tive. Employee turnover is well below average, Lincoln’speople willingly work large amounts of overtime when itis available, and the company actually had to make a rulethat production workers could not be on the factory floormore than a half hour before the start of their shiftsbecause they were so anxious to get to work and startmaking money. The company discovered to its chagrinthat the employees in a number of the acquired foreignoperations had very different tastes than those in theCleveland plant. They did not respond well to the piecerates, they were reluctant to work overtime, and theyshowed, for Lincoln employees, an unusual taste fortaking time off. The acquisitions failed, leading Lincolnto suffer its first losses in a century of doing business.

Organizational architecture can also be used to affectmotivation. For example, creating small business unitscan have strong effects via a number of different mecha-nisms. First, it facilitates measuring performance moreprecisely and so supports giving stronger incentives. Inthis regard, Asea Brown Boveri, the Swiss–Swedish engi-neering company, went to the extreme. In the early 1990sit had over 1,300 separate business units, each a separatecompany with its own balance sheet. These in turn werebroken down into 5,000 profit centers, each of whichaveraged about thirty-five employees and was concernedwith a very narrow product range in a single geographic

Motivation

166

market. Managers in these units were compensated ontheir units’ performance, which was carefully trackedthrough the company’s ABACUS reporting system. Asecond effect of creating such small units is that it letspeople see the impact of their efforts more clearly, whichis itself directly motivating, whether or not pay is tied toperformance. People tend to put more into tasks wherethey perceive they can make a difference and where theresults they generate are clear to them. This enhancementof intrinsic motivation may be especially important whendirect performance pay is not easily employed. As well,organizing in small units reduces the extent of free riding,because the credit for extra results is shared more narrowly.

The basis on which organizational units are defined canalso affect motivation. Most simply, the basis for organiz-ing shapes perceptions of what is important. For example,people in a functional organization are likely to focus onexcellence in their respective functions. The fact that jobsare defined functionally signals that functional excellenceis important to those who designed the firm. The close,frequent contact that naturally occurs with immediatecolleagues, all of whom are in the same function, rein-forces the tendency to see excellence in that function ascentrally important. As well, since evaluations are doneby others in the same function, rewards are more likely toflow to those who demonstrate functional excellence.In contrast, organizing around customer groups orproduction facilities will turn people’s attention in thecorresponding direction.

The recent popularity of the “front-and-back” model fororganization reflects the impact of architecture on motiva-tion. In this model, production and product development

Motivation

167

are organized functionally as the “back-end” of the firm,while “front-end,” market-facing units deal with cus-tomers. The idea is to get functional excellence while pre-serving a customer-oriented focus. Correspondingly, thetwo types of units are measured and rewarded differently.The challenge then is to achieve effective coordinationbetween front and back. In this regard, Tenaris, a globalsupplier of pipe for the petroleum and auto industries, hasa sophisticated supply management system linking its fourcustomer-defined front-end business units operating intwenty countries and the eight manufacturing plantsaround the world that are the back-end of its organization.

The effect of organizational architecture on motivationand behavior also underlies the decision by StanfordUniversity’s Graduate School of Business not to sub-divide its faculty into academic departments, althoughmany other leading business schools do have a depart-mental structure. Instead the Stanford professors areorganized into multiple, overlapping groups for differentpurposes—MBA teaching, doctoral student supervision,faculty recruiting—and the membership of these groupschanges over time. This system is managerially compli-cated, and it would certainly be simpler to organize alongdepartmental lines. The reason it is used is that theSchool wants to foster cross-disciplinary interactions inteaching and research, and it fears that departmental lineswould get in the way of collaboration across fields.

The external boundaries of the firm can also be used toeffect motivation, particularly in a multi-tasking context.For example, it may be important that managers of abusiness both drive current performance and invest tosupport future growth. It is easy to measure current

Motivation

168

performance but probably harder to tell if investments forthe future are appropriate. Then if the business is a unitwithin a larger enterprise, the future may be under-emphasized. (This is especially likely if the managers canexpect to move on to new jobs before the results of anyinvestments become evident.) It may then be better tospin-off the unit and give the managers a large ownershipstake in it, so that they have the more balanced incentivesthat come from the stock price. Of course, the market willnot be likely to do a perfect job of evaluating investmentseither, but investors do have very strong incentives tomeasure performance on this dimension—their personalnet worth is on the line. Thus, their evaluations are likelyto be pretty good and they will certainly enjoy anintegrity that internal measures often lack.

Whether someone should be an employee, usingtools provided by the firm, or an outside contractorowning his own tools may also be responsive to motiva-tional concerns (Milgrom and Roberts 1992: 231–2). It isdesirable that someone using tools at work both producesoutput and maintains the tools. It may be fairly easy tomeasure or accurately infer the effort the worker devotesto production, but telling whether the tools are beingproperly cared for is more difficult—determining theactual depreciation on capital goods is notoriously hard.Thus, there is a precise measure for one activity andonly an imprecise measure for the other. Yet, balancedincentives are clearly needed. Balanced weak incentivescome from making the person an employee, paying ina manner that is largely independent of output, and hav-ing the employer own the tools. To a first approximation,because the worker is paid the same no matter how he

Motivation

169

spends his time, he will be willing to allocate his effortbetween production and maintenance as the firmrequests. If, on the other hand, the worker owns thetools—as an independent contractor—he bears all thecosts and benefits of the effort he devotes to mainte-nance. Then the worker needs to be given comparablystrong incentives for producing output. Providing strongbalanced incentives may most easily be done outside theemployment relationship. Strikingly, contractors muchmore often face explicit performance incentives than doemployees doing similar sorts of work. Which of the twosolutions is best then depends on the ability of theworker to finance the purchase of the tools, his abilityto bear risk, and how important it is to induce high levelsof effort.

Another way that the boundaries of the firm may affectmotivation is through outsourcing. Moving supply to anoutside contractor can reduce influence activities withinthe organization, since there is no longer a common bossover the buyer and seller.

The allocation of decision authority can affect incen-tives as well. Empowering managers may induce them todo a better job of gathering information and makingchoices because they expect that their actions will haveconsequences that they control (Milgrom and Roberts1988b, 1990a; Aghion and Tirole 1997). If, instead, theirdecisions are frequently overruled, this reduces theirmotivation. The cost of this empowerment is that thedecisions that are made are those that are best for the man-ager, not necessarily the firm as a whole, unless the rewardsystems or other mechanisms have brought about align-ment of interests.

Motivation

170

Finally, the financial and ownership structure adoptedfor the firm can affect motivation in many ways. The pub-licly traded corporation gives limited liability to share-holder-owners and allows them to diversify theirholdings, facilitating more risk-taking. Among the risksthat are more easily borne are those of entrusting theactual operations of the business to others, including pro-fessional managers. The downside is that the same provi-sions reduce the incentives for owners to monitor. Incontrast, members of a partnership have very intenseincentives to monitor decision-making in their firms. Themanagement buyouts that became popular in the 1980smeant that agents became principals, and the effect inmany cases was to improve performance radically. Havinga single large shareholder rather than dispersed owner-ship affects the incentives for owners to monitor directorsand managers and so may affect agents’ behavior. Buyingback shares causes a given number of shares held by theexecutives to become a more intense incentive scheme,because the executives now get a bigger fraction of anychanges in the value of the firm.

Processes and routines can also be used to reduceagency problems. An important example is improving theperformance measures. This can be done by investing inthe measurement system to reduce the divergencebetween the measured performance and what the agentactually did, by developing measures that are less mani-pulable, or by adding additional measures that carry extrainformation. In terms of the theory sketched above, any ofthese allows providing stronger incentives.

For example, developing even imperfect indicators ofactual behavior rather than just relying on results can be

Motivation

171

very useful, especially when results are uncertain or comeonly after some long lag. Casino operators spend largeamounts to monitor employees’ and customers’ behaviordirectly rather than relying on any sort of outcome-basedrewards and punishments to thwart cheating. Venturecapitalists gear continued funding of start-ups to thepassing of “milestones,” largely because financial data arenot very informative. Information systems can play animportant role by increasing the accuracy and timelinessof performance measurement and permitting more effec-tive linking of rewards to performance.

As noted earlier, redefining jobs to reduce the need tomulti-task when measures are not of comparable accuracycan also allow giving stronger incentives. More generally,job design can be an important effect on motivation.

Culture can also be a factor in dealing with motivationproblems. Norms about how hard one works, what sort ofrisks one takes, and, generally, about what sorts of beha-vior are appropriate differ tremendously across compa-nies and even among units within them. These norms areenforced by social pressure and the desire to conform. Tothe extent that managers can foster a culture that empha-sizes performance, motivational problems are dimin-ished. Two examples illustrate this.

BP, having disaggregated its exploration and produc-tion stream into small business units, connected theseunits together in peer groups, each of which involvedabout ten business units that faced similar technical andcommercial challenges. Members of a peer group wereencouraged to call upon one another for assistance indealing with such problems. The groups also met fre-quently, without any headquarters personnel present.

Motivation

172

Over time, strong personal networks (and friendships)developed across business unit boundaries, and strongnorms developed about sharing best practice andresponding to requests for help. Remarkable levels ofcross-unit cooperation resulted, despite the fact that theperformance management system did not track or explic-itly reward such behavior.

The sharing of best practice was a crucial element inBP’s successful drive to improve its costs and volumes.Initially the pay of business unit managers was tied to thecost and volume performance of their individual units. Asa strong, shared belief in the importance of cost and vol-ume performance developed and permeated the organi-zation, and as norms developed that units managers woulddeliver the performance they had promised, the explicitlinkage of business unit leaders’ pay and their units’ per-formance was reduced and then eliminated. The valueshad been sufficiently internalized that no explicit incen-tives were needed to motivate managers to deliver perfor-mance, and the compensation system could be directed tomotivating other elements of behavior.

Culture was also crucial to motivation at NokiaCorporation, the Finnish mobile telephone manufac-turer. Nokia nearly failed in the early 1990s. The new topmanagement team who assumed control at that timehelped create a culture where everyone was dedicated tomaking the company succeed. The culture encouragedindividual initiative, hard work, and judicious risk-takingin a technologically turbulent environment where it wascrucial that lower- and middle-level people acted quicklyand decisively on the basis of the information they alonehad. Successes were celebrated, and failures were not

Motivation

173

punished. Even when problems in managing logistics ledto a 50 percent drop in the stock price in 1995, no one wasfired. An atmosphere of mutual trust was establishedamong employees and between them and top manage-ment. Fear, so often a central fact of corporate life, wasbanished. People were simply expected to do their bestand were trusted to act in the best interests of thecompany. They responded by doing just that.

Nokia became one of the most successful corporations ofthe 1990s, at one point becoming Europe’s most valuablecompany, and its success continued even after the burstingof the bubble in the telecommunications industry. It con-sistently introduced great new products while achievinglower costs than its competitors, in large part throughhighly disciplined operations. Yet, all this was done ina context where pay was not high relative to the employees’outside options—a top executive described Nokia’s peopleas “happily badly paid”—and where rewards were tied togroup or overall corporate performance.

The model of “high commitment human resourcemanagement systems” presents a general example of usingseveral different elements of PARC to achieve motiva-tion. As outlined by the leading text in the field (Baronand Kreps 1999: 190), key elements of such systemsinclude guaranteed employment (except after egregiousmisbehavior); egalitarian values and norms; self-managedteams for organizing production; attempts to make workinteresting and fulfilling; premium compensation, per-haps involving team, unit, or firm-wide (but not indi-vidual) performance pay; rigorous pre-employmentscreening and extensive socialization and training ofemployees; transparency of information within the firm

Motivation

174

and open communication channels between employeesand managers; a strong culture focused on an overarchinggoal such as the organization’s vision; and a strongemphasis on employees’ symbolic and financial “owner-ship” of the firm. The basic deal is that people work hardand cleverly, in the interests of the firm, in return forgood pay, empowerment, trust, and interesting, fulfillingwork. They do so first without any obvious, explicit for-mal incentives because they identify their interests andthose of the firm more closely than in the set-up assumedin the agency models. Beyond that, mutual monitoringand social pressure among employees enforce the desiredbehaviors. The screening, socialization, and identifica-tion with the vision help ensure that the enforcementmechanisms are at work.14

Such systems are especially attractive when it is difficultto measure performance accurately, and thus to provideincentives through more direct reward systems. There arecosts to establishing a high commitment system, and manyof these costs are fixed and sunk. Pay is above the mini-mum necessary to attract and retain workers, screeningand training are costly, creating an atmosphere of trust isnot easy or automatic, and getting the workers and man-agers to buy into the system may be difficult. It is worth-while if it results in sufficiently higher levels of effort thanan explicit performance pay system generates. This willtend to be the case when effort is important but perfor-mance measures are bad.

The trick in such systems is to ensure that the standardsdo not degrade, so that effort is low but the costs remainhigh. Perhaps the surest way for this to happen is for theworkers to lose trust in management because it appears

Motivation

175

either to have violated the terms of the deal or to nottrust the workers. This example points to an apparentcontradiction with the use of measures and rewards.Closely monitoring behavior and explicitly rewarding it,particularly financially, may undermine trust. The near-universal norm of reciprocity can lead to workers’performing at quite high levels if they perceive that theyare trusted and that the firm is treating them well. Thenorm, however, may not survive if the workers are closelymonitored and if the firm introduces explicit incentivepay. Aspects of culture and explicit performance incen-tives may be substitutes, at least over certain ranges. Inthis case, if the culture supports high levels of effortprovision, it may be wise not to use explicit incentive pay.Certainly, a small amount of incentive pay supported bythe measurement systems it requires will not be worth-while if it undercuts the culture of trust and the normof reciprocity.

More generally, it is important to consider the interac-tion of different mechanisms in affecting motivation.While some elements of organization design may be com-plementary in motivating people, others may conflict andundercut one another. Performance-based rewards aremore effective if performance is measured better.Creating small business units works better if their man-agers are genuinely empowered. On the other hand,overly close monitoring of behavior against mandatedprocesses and procedures may destroy the motivatingeffects of empowerment. We will return to the problemsof designing motivation systems in the remaining chap-ters, where we consider more specific objectives that thedesigner is attempting to realize.

Motivation

176

Notes

1. For surveys of this subject, see Gibbons (1997), Gibbonsand Waldman (1999), and Prendergast (1999).

2. While motivation problems can be manifested at the grouplevel, we focus our discussion on the problems of motivat-ing individuals because these are logically prior andbecause most of the arguments are more directly made inthis context.

3. See Avery, Chevalier, and Schaefer (1998) for an analysis ofCEOs’ motivations for making acquisitions.

4. This is not to say that reputational concerns cannot helphere. However, the fact that behavior is unobserved willmean some unavoidable inefficiency.

5. The basic reference for simple agency theory isHolmström (1979). See also Hart and Holmström (1987)for a survey and Milgrom and Roberts (1992: chapters 5and 7) for an elementary formal development.

6. The simplest case is one where the agent’s effort is eitherhigh (“working hard”) or low (“being lazy”) and the prob-abilities of different outcomes are shifted by the effortchoice. In fact, most of the insights obtained in more gen-eral formulations arise in this context (Hart andHolmström 1987). Another central case is one where theagent’s preferences show constant absolute risk-aversionwith respect to his money income less the cost of effort (sothere are no income effects), performance varies directlywith the effort choice, the additive noise term in the per-formance measure is normally distributed, and the contractis linear (Holmström and Milgrom 1991; see also Milgromand Roberts 1992: chapter 7). This set-up gives very richpredictions/explanations and our discussion is largelybased on this case.

Motivation

177

7. The assumption of risk-neutrality for the principal is areasonable approximation when, for example, we thinkabout the compensation of an employee by a large firmowned by shareholders who are able to diversify theirportfolios and thus are largely indifferent to the variationin the firm’s returns that one person’s pay represents.

8. The optimal strength of the incentives (in the linear casedescribed in note 6) is directly proportional to theexpected marginal benefit to the principal of extra effortand inversely proportional to a term that is equal to oneplus the product of three terms: The variance of the mea-surement error, the agent’s risk-aversion parameter, and aninverse measure of how the agent’s effort choice respondsto increased incentives. The claimed results follow fromthis relationship. For details, see Milgrom and Roberts(1992: chapter 7).

9. The dependence is actually on the variability of pay, whichdepends on the intensity of incentives. But the agent’soptimal effort choice will cause the marginal cost of effortto be equated to the marginal incentive intensity, so thatthe risk cost can be expressed in terms of the marginal costof effort.

10. Unless the monitoring is certain to catch any misbehavior,the amount paid will necessarily be above the level of theemployee’s best outside option, so that being fired is costlyand the threat of termination is a real incentive. The extrapay is called an “efficiency wage” because paying morethan the employee’s opportunity cost increases effort andefficiency.

11. San Jose Mercury News, October 2, 2000.

12. It is useful to consider the extreme case where the agent’scost of effort depends only on the total amount that theagent provides, not on its allocation among tasks, the

Motivation

178

marginal cost of effort is increasing linearly in the effortlevel, and pay is a linear function of the measured perfor-mance on each activity. If multi-tasking is induced, theintensity of incentives must be the same for both activities.In this case, the optimal linear payment scheme inducingmulti-tasking gives strictly weaker incentives than would beprovided for either task in isolation. Thus, the total effort isstrictly lower under multi-tasking than if effort is inducedon just one activity, even the poorly measured one.

13. The practice of paying for individual or small-group per-formance with shares of company stock is not subject tothis criticism, which is aimed at situations where employ-ees are given stock in amounts that are independent oftheir current performance with the idea that holding theseshares will motivate effort.

14. One interpretation of these systems is they involve an effi-ciency wage (see note 10 in this chapter) and the workerssupply lots of effort because if they collectively do not, thefirm will revert to more standard Human ResourceManagement systems that are less favorable to the workers.

Motivation

179

180

5

Organizing for Performance

In this chapter we seek to apply the concepts and theoriesdeveloped in earlier chapters to examine how firms canorganize themselves to increase their performance againsttheir current strategies. The key idea is to design theorganization to provide as focused, intense incentives aspossible within the constraints implied by the corporateform and the interdependencies that it both creates andis meant to control. Doing so necessitates a variety ofchoices of architecture and routines, supported by cul-tural changes, that together can be called disaggregation.The key architectural elements involve redrawing thehorizontal and vertical boundaries of the firm to increasestrategic focus; creating relatively small subunits withinthe organization in which significant decision rights arelodged; and decreasing the number of layers of manage-ment and the extent of central staff. Routines andprocesses are altered to hold the subunits accountable fordelivering performance while linking them together byvarious means to manage the interdependencies amongthem. Finally, cultural norms are developed that facilitatethe pursuit and realization of improved performance.1


181

In fact, some firms have long employed organizationaldesigns that involve many of the features of the looselycoupled, disaggregated model. Johnson & Johnson, thepharmaceutical, medical equipment, and consumer prod-ucts company, is a prime example (Pearson and Hurstak1992). Johnson & Johnson is composed of over 150 sepa-rate companies that each serve different markets. Thefiercely independent companies carry out their own prod-uct development, production, marketing, and sales, pay-ing dividends to the parent. The highly decentralizeddesign generates intense incentives in the companies tocreate new products and increase sales. Meanwhile, astrongly shared set of values and norms helps ensure thatbehavior is aligned with overall performance.

In the last decade, however, many more firms haveadopted this design model in response to increased needsto improve performance. Falling barriers to internationaltrade and investment and the increased ease of long-distance communication and transportation have allowedfirms to enter new foreign markets. This has increasedcompetition in product and service markets aroundthe world, making improved efficiency necessary for suc-cess and even survival. At the same time, capital marketsare arguably becoming more demanding, puttingincreased pressure on firms to perform. The increasedacceptance of creating shareholder value as the primeobligation of management, the greater activism of insti-tutional investors, and, in some parts of Europe, theemergence of hostile takeover bids all contribute to this.Finally, the increased linkage of top executives’ com-pensation to performance, which has gone furthest in the United States but has occurred elsewhere too, means


182

that the incentives to improve performance are muchstronger than in the past.

BP plc, the integrated oil and natural gas companyformerly known as British Petroleum, exemplifies use ofthe disaggregated model to improve performance. Althoughwe have discussed aspects of BP’s organization at a num-ber of points already, a more systematic treatment isworthwhile.

In recent years, BP has reported some of the highestprofits ever recorded by any corporation—over $14 billionin 2000. While high prices for crude oil certainly con-tributed to BP’s outstanding financial results in thisperiod, the company’s extremely cost-efficient operationswere crucial to its underlying performance, which wasstrong throughout the last part of the 1990s and into thiscentury. Indeed, BP was widely recognized in the energyindustry for its effectiveness in finding hydrocarbondeposits and its efficiency in bringing these into produc-tion and extracting the crude oil and natural gas. Thisoperational excellence was actually maintained andimproved in quite difficult circumstances that could haveresulted in major disruptions in the workings of the firm:Between 1998 and 2000, the company absorbed two othermajor integrated oil companies, Amoco and Arco, thattogether were almost as large as BP itself, as well as a lubri-cants company, Burmah Castrol, a retail joint venture ithad in Europe with Mobil, and Veba’s German gasolineretailing operations.

Strategic and organizational changes implementedover the preceding decade under three successive CEOsformed the basis for BP’s success. A dozen years earlier,BP had been a politicized, top-heavy bureaucracy

organized through a cumbersome matrix structure. Thecompany was still spread across numerous distinct linesof business, the result of its not having yet completelyundone the conglomerate diversification in which it(along with the other major oil companies) had indulgedin the 1970s. Financial proposals required fifteen differ-ent signatures before they could be accepted; head-officestaff filled a thirty-two storey building; and meetings ofeighty-six different committees absorbed the top execu-tives’ days. Performance was in decline, the company washeavily indebted, and in 1992 it reached a financial crisisthat almost resulted in its bankruptcy.

The changes that would eventually transform BP beganwith its divesting unrelated lines of business. Thisprocess, begun in the 1980s, was completed by the early1990s. The company was then focused on three basicbusinesses or streams: Upstream oil and gas explorationand production; downstream petroleum refining andmarketing; and petrochemicals. These lines of businesswere obviously related, as upstream’s products are thebasic inputs to the other two streams’ production. In fact,however, they could be and were run quite separately in alargely decoupled fashion. Well-functioning world mar-kets allowed efficient purchase and sale of crude oil, so thecompany did not need to rely on internal transactions.

The organizational changes began under CEO RobertHorton, who took over in 1989. His “Project 1990” soughtto improve the speed and effectiveness of managerialdecision-making. Horton transferred authority for manydecisions from the corporate center to the business streams.In the process, layers of management were eliminated andheadquarters employment was reduced by over 80 percent.


183

Employees were encouraged to take responsibility andexercise initiative, and values of caring, trust, openness,teamwork, and cooperation were espoused. At the sametime, as economic difficulties mounted, capital budgetswere slashed and employment was cut deeply.

Horton’s abrasive personal style and the dissonancebetween the proclaimed values and the reality of job cutsalienated employees, while performance continued todeteriorate in the context of the general economic slow-down of the early 1990s. In 1992 Horton was replaced asCEO by David Simon, who had been Chief OperatingOfficer. Simon was more popular among the BP employ-ees, but he continued Horton’s aggressive rationalizationagenda. Employment fell from more than 97,000 in 1992to just over 50,000 in 1995. Some of this was associatedwith divestitures and asset sales, but a significant amountof it represented reductions in employment in ongoingoperations. Performance improved radically. The com-pany moved from a loss of $811 million in 1992 to a profitof $2.4 billion two years later, while debt levels fell by$4 billion.

The biggest changes during this period occurredwithin the upstream, exploration and production business,BP Exploration (BPX). There, John Browne, who wouldsucceed Simon as BP Group CEO in 1995, undertook afundamental organizational redesign. The model, whichBPX called an “asset federation,” typified disaggregateddesign. It later was applied across the company as a whole.

Browne began by refocusing the upstream strategy onfinding and exploiting large hydrocarbon deposits wherethe technical difficulties and attendant risks meant thatBP’s expertise and size gave it a relative advantage over


184

smaller firms. (In this period, smaller exploration andproduction companies—labeled “petropreneurs”—wereoften more successful than the large integrated oil com-panies.) This led to disposal of a number of smaller hold-ings, purchases of other fields that complementedexisting ones, and the redirection of exploration towardsareas that were more likely to have really major deposits.This meant that each of the company’s oil or gas fieldswas of some substantial size. In the late 1980s these fieldswere concentrated in the North Slope of Alaska and theNorth Sea, whence BP had drawn most of its productionsince the 1970s. However, focusing on major opportuni-ties meant that in the future BPX’s exploration activitieswould be focused on new areas where technical or politi-cal difficulties had prevented earlier development. Thus,BPX’s operations would become increasingly dispersedaround the globe and would be increasingly in developingand transition economies.

The next move was organizational. BPX had been struc-tured through a collection of geographically definedRegional Operating Companies (ROCs), which had staffsof technical and business people overseeing the actualoperations. The heads of the ROCs and of the functionsjoined Browne in a Global Management Group that ranthe stream. Performance data were normally aggregated tothe level of the ROCs, and the managers of the actualfields had very limited discretion and very little controlover the resources used in their units. Browne, however,began to push performance evaluation discussions downto the level of the individual fields. This led to a consciousexperiment in organizational design, with the managers ofa number of fields being given authority to decide how to


185

run their operations and how to meet performance targetsthat were negotiated directly with top management ofBPX. When this change resulted in increased outputs andreduced costs, the model was applied throughout BPX,beginning in the crisis year of 1992.

All the exploration and production operations weredivided into some forty separate business units, calledassets, each of which consisted of a major oil or gas fieldor a group of co-located fields. Each was headed by anasset manager (later called a business unit leader). TheROCs were eliminated, with senior management of thestream being pared down to Browne and two others. Asthe BPX Executive Committee (ExCo), they togetherdirectly oversaw the assets, with no intermediate layers ofmanagers. The technical and functional staffs were alsolargely dispersed to the assets.

The asset managers were given charters that set boundson their activities (e.g. limiting their drilling to their ownsites). They also signed explicit, individual performancecontracts with the ExCo, agreeing to deliver specified lev-els of performance in terms of production volumes, costs,and capital expenditure. Within their charter bounds andthe limits of general corporate policy, the asset managerswere then empowered to figure out how to achieve theirpromised performance. They could decide on outsourc-ing and choose suppliers, do their own hiring, and deter-mine where and how to drill.

The performance of individual assets was not aggre-gated below the level of the stream itself and was fullytransparent to Browne and the ExCo members. Theytracked performance closely, especially through rigorousQuarterly Performance Reviews. Through conversations


186

in these meetings, Browne coached the asset managers,helping them develop their managerial skills andinculcating the values and norms he sought to spreadthrough BPX.

The performance contracting was not limited to theasset managers. Instead, the promises made in the perfor-mance contracts at the asset level became the basis forperformance contracts for all the individuals within theasset. All employees’ compensation was tied to theirassets’ performance and to the overall performance ofthe stream. This increased the variability of pay and theintensity of incentives significantly.

The asset managers found the new system very liberat-ing, but the leanness at the top meant they could not rely onstream headquarters to advise and support them when tech-nical or business difficulties arose. To respond to this need,the assets were aligned into four peer groups that (aftersome experimentation) were defined on the basis of the life-stage of the assets. One group consisted of actual explo-ration activities, including obtaining rights to develop fields;the second included assets that were being developed andbrought into production; the assets in the third group werein full, plateau production; and the fourth included assetsthat were approaching the end of their economic viabilityand were in decline. The key point was that assets within agroup, although geographically dispersed, were likely toface similar technical and commercial problems.

The asset managers were encouraged to rely on thepeer group colleagues for support. Indeed, the peergroups were designed to facilitate mutual assistanceamong their member assets and to promote the sharing ofbest practice. To this end, the system of peer assists was


187

established under which an asset facing a problem couldcall on people from other assets to come and help solvethe problem. As well, numerous other “federal groups”came into being, linking people with common interestsand challenges across the different assets.

The peer groups were also given another role early on,called peer challenge. Under it, peer group memberswere expected to challenge one another on the targets thatthey negotiated individually with the ExCo. This processallowed the asset managers’ collective expert knowledgeto be brought to bear in establishing targets. Later, thepeer groups each took collective responsibility for meet-ing the performance targets of the member assets and forallocating capital among them.

At the same time, increasing reliance was put on out-sourcing. This included significant elements of thehuman resource management and accounting functionsand extended even to activities previously seen as criti-cal, including the generation of seismic data on potentialnew fields (only the interpretation of the data was keptin-house). Strikingly, the logic of the performance con-tracts was sometimes extended to outside suppliers,whose payments were made a function of their perfor-mance. A triumph for this approach was the Andrewfield in the North Sea, which had previously beenbelieved to be too expensive to bring into production.By sharing cost savings with its contractors, BP was ableto develop the field at a fraction of the original costestimates and in much less time than had been believednecessary.

This organizational model led to BP’s remarkable suc-cesses. New fields were found and developed, many in


188

areas that were previously thought to be technically toodifficult to be economically feasible. The cost of develop-ing fields was reduced substantially and kept beingsqueezed, and the productive life of assets was extendedlong beyond what had been believed possible.

After Browne became CEO of British Petroleum in1995, this model was applied across the whole of thecompany. The appropriate definition of assets was lessobvious in the other streams than it had been in BPX,and establishing the right performance measures alsopresented some challenges. Still, the system of discretebusiness units, peer groups and peer assists, smallExecutive Committees for each stream, performancecontracts, and peer challenge was instituted. Performancemanagement began with a performance contract betweenCEO Browne and the Board, which then cascaded downthrough contracts for the Managing Directors who ledeach stream through to the Business Unit Leaders andthen on to the individual employees. In addition, a set ofRegional Presidents was established to provide assurancethat company policies were being met with regard toissues that tended to cross stream boundaries and werebest handled on a national or regional basis, such assafety and environmental matters or legal and regulatoryaffairs.

These changes in the architecture and routines eventu-ally led to fundamental cultural changes. BP’s peopledeveloped a deep, intrinsic dedication to delivering ever-improving performance. Strong norms emerged ofmutual trust, of admitting early when one faced difficul-ties (“no surprises”) and seeking assistance when needed,of responding positively to requests for help, of keeping


189

promises about performance. These had powerful effects,generating a remarkable amount of cooperation while stillinducing great initiative.

This model has been adjusted since, but in its basiclogic persisted through the 1990s. It proved especiallyvaluable during the absorption of Amoco and Arco, wherethe newly acquired assets were quickly integrated into theBP system with the support of the peer groups.

In the wake of BP’s success, many firms have soughtexplicitly to emulate its organizational design. Othershave adopted similar organizational designs on their own.There is, of course, no one best way to organize that allfirms should adopt. The best organization designdepends on the strategy being pursued, the market andnon-market context, and the administrative heritage of theorganization. Nonetheless, real gains in performance canoften be achieved by adopting designs that adhere to thebasic logic underlying BP’s disaggregated model. Theseinclude a tendency towards focusing the activities of thefirm to a select set; creating business units with clearscope of responsibility and clear accountability; givingstrong incentives for unit performance; linking units hor-izontally rather than requiring all communication to passup and down through the hierarchy; flattening the hier-archy and increasing spans of control; outsourcing;improved information, measurement and communicationsystems; and, ultimately, the creation of a culture that isoriented to delivering performance.

These design elements are complementary and so thesignificance and impact of each can be understood onlyby recognizing the interactions among them all. However,it is worth first looking at each in isolation.


190

Vertical Scope

An important element of designing the organization forgreater performance is to focus the firm on just thoseactivities where it can create the most value. For manyfirms, this has involved vertical disintegration, turningover to others the provision of goods and services that thefirm formerly provided for itself.

A striking manifestation of this process of outsourcingand vertical disintegration is a set of firms exemplified byNike (Brady and de Verdier 1998; Whang and de Verdier1998) and Benetton (Stevenson, Martinez, and Jarillo1989) that have adopted the role of “vertical architect” or“value chain organizer.” This role involves the lead firm’sorganizing and managing a value chain—in sports shoesand, more recently, apparel and sports equipment atNike, and in fashion wear for Benetton—but actuallyowning few of the assets involved and carrying out few ofthe activities that are needed to create value. Nike, forexample, outsources all its production, but does the prod-uct design, marketing, and distribution to (independent)retailers. Benetton actually outsources the basic designwork and most of the manufacturing for its products. Italso relies on retail outlets that are independently owned,although selling only Benetton products, and it deals withthese retailers through agents who are not Benettonemployees. The outlets are, however, linked to Benetton’sinformation systems to track sales. Benetton takes care ofcreating the patterns from the designers’ drawings, dyesthe clothing, handles the logistics of distribution, andruns the advertising and marketing for the brand. In bothcases, the lead firm manages a complex set of relations


191

with other value chain participants and coordinates activityamong them.

This model has been extensively adopted in the elec-tronics industries. For example, in the personal computerindustry, many of the leading firms outsource almostall manufacturing to such electronic manufacturing ser-vices (EMS) companies as Solectron and Flextronic. Bothof these EMS companies do tens of billions of dollars ofbusiness a year, but they have no products of their own.Further, the computer firms are also beginning to out-source logistics, order fulfillment, and post-sales service,and even the design and manufacture of their low-endproducts.

There are, of course, numerous very good reasons whya firm should prefer to purchase goods and services ratherthan provide them in-house.2 Fundamentally, unless thefirm has some special competence in carrying out thesupply activity, others are likely able to do the job betterand more cheaply. This could be because they specializein this task and, through learning, become more profi-cient at it; or because they enjoy economies of scale insupplying multiple customers which the buying companycannot itself realize; or because their focus reduces orga-nizational complexity relative to the integrated alterna-tive and thus results in lower costs of management.Greater focus also reduces measurement and attributionproblems, facilitating the provision of stronger incentivesto employees. Moreover, relying on competition to setprices may be much better than attempting to determineinternal transfer prices. Outside supply is also attractivebecause it is probably easier to induce competition amongexternal suppliers than it is when the suppliers are inside


192

the buyer’s firm. In particular, replacing an unsatisfactoryoutside supplier is much easier than getting rid of aninternal supplier that is not performing well.

On the other hand, a variety of reasons have beenoffered to justify vertical integration. Some of these standup to rigorous analysis; others do not. For example, thetransaction cost and property rights schools of econo-mists accentuate the protection of specific assets againsthold-up. Owning the assets associated with providing thegood or service and supplying it in-house may providebetter protection for such assets and thus stronger invest-ment incentives. Other reasons to integrate have todo with the protection of intellectual property or, moregenerally, the difficulties of operating efficient marketsfor information. Relying on internal provision may pro-tect sensitive knowledge from misappropriation by sup-pliers. Because interests may be more aligned within thefirm than across firm boundaries, internal provision mayalso facilitate the transfer of valuable informationbetween the supplier and customer.

The provision of appropriately balanced incentives formultiple activities can also drive firm boundaries. Theexample discussed in Chapter 3 of the choice between anoutside distributor and an employee sales force is illustra-tive. If it is important that sales people also serve as aninformation conduit from customers back to productdevelopment, then keep sales inside the firm. It will bedifficult to motivate an outside distributor to undertakethis activity, which is hard to measure and thus to reward,while also providing incentives for sales that are suffi-ciently strong that the distributor will focus on the firm’sbusiness rather than that of other firms it might represent.


193

Instead, the firm should employ its own internal salesforce. It would then give the sales people incentives thatare balanced (to induce both sorts of activity) but rela-tively muted (since strong incentives for the ill-measuredactivity are likely to be too random and costly) and simplytell them they cannot sell other firms’ products. In arelated case, integrating forward into retail sales mayfacilitate the provision of customer education about use ofthe product. An independent seller, unless it has an exclu-sive territory, may be reluctant to provide this costly ser-vice. It fears the customer will come to it for theinformation but then actually buy from a lower-pricedcompetitor that free-rides on the first seller’s educationalactivities.

Yet another reason to favor internal procurement maybe to lessen the incentives for opportunistic slacking onquality (although in fact it seems that many managersview internal suppliers’ quality as more problematic thanoutsiders’). Finally, to the extent that being in the samefirm, with the managers of both activities reporting to asingle boss and with the employees all belonging to a sin-gle company, facilitates coordination, then integration isfavored.

To the extent that outsourcing involves trading offlower costs of production against increased transactionscosts (from hold-up, information leakage, or whatever),globalization, improved information and communicationtechnologies, and more flexible manufacturing systemswould favor a shift to more outsourcing. The Internetfacilitates finding new suppliers, improved communica-tions makes dealing with them easier, and lower transportcosts and reduced trade barriers allow working with more


194

distant suppliers. Thus, the costs available throughoutsourcing should be lower, and this favors doing moreof it. More flexible manufacturing systems, such as com-puter numeric controlled machines, also favor increasedoutsourcing by reducing the threat of hold-up: The sup-plier’s flexible equipment can be reassigned to other usesif the buyer tries a hold-up, and the flexibility of competingsuppliers means that the buyer can find others with whichto deal if the supplier tries to grab more than its share.Thus, standard theories of the make–buy choice wouldpredict increased outsourcing in the current context.

All of these theories, however, have largely focused onone-off dealings between discrete parties who do not actas if they expect to transact again and who do not worryabout the effect of their behavior in this transaction ontheir reputations. They can then be assumed to maximizetheir immediate returns in the current transaction.3 Insuch a context, it is relatively easy to justify integration,because of the fear of excessively sharp dealing whentransacting with outside parties.

The distinguishing feature of much of the outsourcingthat companies have been adopting recently, however, isprecisely that the supplier–buyer interaction is structuredas an ongoing relationship, partnership, or alliance. Thelong-term nature of the relation radically changes theincentives that can be provided, whether these are tomake investments, respect intellectual property, maintainhard-to-monitor quality, or whatever. They therebygreatly increase the attractiveness of outsourcing com-pared to a situation where the outside dealings are simple,one-off, arm’s length ones. Repeated dealings allow formuch more cooperation. Meanwhile, moving the supply


195

relation from being a long-term one within the firm toanother long-term one that crosses firm boundaries per-mits gaining many of the advantages that were alwaysinherent in outside supply. The opportunities for induc-ing initiative are also increased.

In terms of the cooperation–initiative trade-off, orga-nizing through a long-term relationship is an organiza-tional innovation that shifts the frontier of feasiblecombinations to allow achieving much more cooperationthan one-off market dealings but much more initiativethan internal supply. Points like R are now available thatwere not before (Figure 9).

The leading Japanese automobile manufacturers—especially Toyota—were among the first to develop fullythe model for this sort of long-term supplier relations.4 Inpart the Japanese firms chose to outsource extensively inresponse to government policies that favored dealings withsmaller firms. In part too, the permanent employment


196

Cooperation

Initi

ativ

e

Marketdealings

Internalsupply

R

Figure 9. Long-term relations expand the set of attainablebehaviors

policies adopted by the major Japanese firms may havefavored using outside suppliers to maintain flexibility.Still, the model proved very effective: Its use has beencredited with a significant portion of the cost and qualityadvantages the Japanese auto firms enjoyed over their U.S.rivals in the 1970s and 1980s (Womack, Jones, and Roos1990). Since then, other auto companies have adopted thismodel, and it has spread widely to other industries.

In such an ongoing relationship, the relative focus foreach party is supposed to shift away from attempting toappropriate value (at the possible cost of reducing thetotal value created) and towards creating value, with thepresumption that the extra value will be shared appropri-ately. The basic logic of the relation is that incentives forcooperative, value-enhancing current behavior are providedby the promise that good behavior will result in continuedfuture cooperation and sharing of the resultant returns,while misbehavior will result in future punishments, upto and including the termination of the relationship. Theparties in effect enter a “relational contract”—a sharedunderstanding that they will cooperate and divide theresulting gains. The exact behavior that will be needed inevery eventuality cannot reasonably be described in a formalcontract and verified by the courts, so the relational contractmust be self-enforcing. The enforcement mechanism is thefuture behavioral responses of the parties to current good orbad behavior.5 To make this logic work, a number of condi-tions must be met. These flow from the fundamental incen-tive problem and the nature of its solution.

The fundamental issue is that, at various points in time,either of the parties may see opportunities to increase itscurrent returns by behavior that hurts the other party but


197

that cannot be effectively deterred through normal,court-enforced contracts. For example, the suppliermight not expend as much effort as would be efficient atcontrolling costs or maintaining hard-to-observe quality.Similarly, the buyer might refuse to absorb its share of thecosts that the supplier has incurred to serve the buyer’sneeds, even though the understanding between the par-ties was that the buyer would compensate the seller. Orthe buyer might make unreasonable demands for speedierdelivery and threaten not to pay unless the seller acqui-esces. If cooperation is to be maintained in the face ofsuch temptations, then giving in to temptation todaymust bring consequences in the future that are suffi-ciently unpleasant that today’s gains are more than offsetby tomorrow’s losses.

In effect, each party must always anticipate that resist-ing temptation will result in continued cooperation whilesuccumbing will elicit punishment—a response from theother party that is worse for the tempted party than on-going cooperation. Further, the difference in the presentvalue of the two possible streams of future returns mustexceed the magnitude of the immediate gain realized bygiving in to the temptation. When this is so, the overallreturn to cheating is negative and so misbehavior isdeterred.

Thus, a first requirement for a successful partnership isthat there must be an opportunity to create value by coop-eration, over and above what would be available undereither self-supply by the buyer or normal, short-termmarket transactions. This immediately means that there islikely to be little reason for adopting a relationship-basedmodel if there are numerous external suppliers who are


198

always ready to meet the buyer’s needs efficiently. Thiswould often seem to be the case when procuring stan-dardized commodities. It also applies even when thepurchased item is not standardized but producing it doesnot involve extensive specialized investments in knowl-edge or physical capital.

The long-term outsourcing arrangement must thenensure that this gain is shared between the parties so thateach is in fact better off in the relationship than out of it.

Further, for the threat of punishment to be effective, itmust be that the injured party can retaliate and do thingsthat are relatively unpleasant for the transgressor. Endingthe relationship, thereby depriving the cheater of thefuture profits it would have earned, is often the strongestpunishment available. An alternative might be revertingto the short-run, non-cooperative mode of market deal-ing. Moreover, these options must be sufficiently attrac-tive for the punishing party that it will prefer to adoptthem than to continue to deal with the transgressor. Sinceconjectured punishments that would hurt the punishertoo much will not likely be imposed, they cannot determisbehavior.

Note that the larger the gains from cooperation overalternative arrangements and the worse the punishmentsthat can be imposed, the easier it will be to induce coop-eration. So if the parties are really committed to the rela-tionship and do not fear that a change in circumstancesmight make them both want to end it, they may actuallygain by worsening their outside options and makingthemselves more dependent on one another. For example,the seller might close its marketing group so that it is in aworse position if the relationship ended, or the buyer


199

might allow its capabilities in manufacturing the item inquestion to erode. Of course, this is a double-edgedsword—it helps as long as the relationship continues, butcan be very harmful if it dissolves.

Further, it is important that the participants placeenough weight on their future dealings. This is facilitatedif the dealings are more frequent and if the parties do notheavily discount future returns. Clearly too, it is impor-tant that the relationship be expected to continue more orless indefinitely, or else, as the end approaches, there isnot enough future in which to give rewards and punish-ments to deter giving in to temptation, and cooperationwill tend to break down.

Finally, it is better if it is clear whether a party actuallyhas transgressed, so that punishment is meted out if andonly if it is warranted. Otherwise, cheating may not bepunished, and so cannot be as effectively deterred, orpunishments may be imposed when they are notdeserved, limiting the extent of realized cooperation andmutual gain (Abreu, Pearce, and Stacchetti 1990).

Features of the way that leading firms actually arrangetheir long-term supplier relations are responsive to theserequirements.

First, long-term relations are not used for acquiring allproducts. When the product or service is standardized,when numerous suppliers are available, and when deliveryand quality are not problematic, then simple arm’s lengthdealings are used. This is because there is no real basis forcreating value through a partnership over and above whatis possible in regular market arrangements. Also, someactivities are systematically kept within the firm rather thanoutsourced to a partner. In the auto business these typically


200

include at least product development and assembly,although there have been recent experiments in turningassembly over to partners. Almost all firms do the careermanagement aspects of the human resource function andthe management of their financial resources themselves.Arguably, both the knowledge needed to carry out these activities and that gained in doing them are so crucialto the firm that there is no value in having another under-take them.

Second, significant attention is given to selecting part-ners and establishing the basic understandings with them.For example, in selecting suppliers for Toyota’s first whollyowned assembly operation in the United States, Toyota’spurchasing staff spent months meeting with each of thesuppliers it considered, and more time with the ones itactually chose, visiting their facilities, encouraging them tovisit Toyota’s, teaching the Toyota Production System, andestablishing common expectations (Milgrom and Roberts1993). This was true even of firms that were already sup-plying NUMMI, the joint venture with GM in Californiathat Toyota managed. This intense and extended inter-action was not aimed at developing and negotiating anydetailed, explicit contract. Indeed, the actual contractswith the suppliers were short and very simple, essentiallycommitting the partners to work together to resolve prob-lems as they arose. Rather, it was first a screening process,as Toyota sought to identify suppliers with whom the mostvalue could be created and who would be cooperative part-ners. Then it was a matter of establishing the relationshipand the shared understandings that it entailed.

Once the relationship is established, the gains fromcooperation between the auto company and the supplier


201

are in fact shared, with both parties getting more thanthey would in their next-best alternatives. Prices are nego-tiated. Toyota then helps its suppliers improve their effi-ciency, even assigning personnel to the suppliers’ factoriesto serve as consultants, and cost improvements do notimmediately result in offsetting price reductions. Thus,both parties get some benefit from the cost improvements.

The parties also pay attention to the punishments thatcan be inflicted for bad behavior. The most direct, extremeform of punishment is to sever the relationship. This isobviously easier for the buyer—and thus a more crediblethreatened punishment—if there are other qualified sup-pliers to which it can turn. Toyota in fact maintains a“two-supplier” policy: While any particular part or com-ponent for a specific model of vehicle will have a singlesupplier for the life of the model (roughly four years), therewill normally be another firm supplying similar inputs forother models. For example, one firm might supply head-light systems for the Camry, but another would supplythem for the Corolla. If necessary, Toyota will take on thissecond-source role itself. For example, Toyota has longrelied very heavily on Denso for automobile electricalcomponents and systems. As electronics came to becomemore and more important in cars, Denso developed majorcapabilities in the area, and Toyota began to be uncom-fortably dependent on this one supplier. It then chose todevelop significant internal capabilities in electronics. Itthus is at least conceptually possible to cease dealing withDenso, thereby providing a severe but plausible punish-ment threat (Helper, MacDuffie, and Sabel 1998).

The two-supplier policy in fact permits giving morenuanced incentives than simply the threat of termination.


202

Toyota keeps careful track of its suppliers’ performancein cutting costs, improving quality, coming up with ideas,and behaving cooperatively. Suppliers are rated using thisinformation, and when models are redesigned, higher-rated suppliers get more business. This induces somecompetition between the suppliers, and this competitivepressure is a source of the advantage of outsourcing over(typically monopolized) internal supply. Meanwhile,compared to arm’s length, short-run dealings, thestronger incentives that can be provided for mutuallyadvantageous, value-creating behavior in an ongoing rela-tionship also favor this approach.

There are costs to the two-supplier policy. Two rela-tionships must be managed, rather than just one, and thistakes time. There may be some loss of possible economiesof scale. More subtly, there is less value being createdbetween Toyota and any one of its partners than therewould be if it were a sole-supplier relationship. This, inturn, may limit the extent to which the promise of futurecooperation and value sharing can induce good currentbehavior. There is, thus, a trade-off here: Having only onesupplier means that more value may be created if cooper-ation is maintained, but having multiple suppliers may, byincreasing the range and credibility of punishments, makea higher level of cooperation more likely to be achieved.

A major difficulty in managing a system based onimplicit threats and promises rather than formal con-tracts is in knowing whether the parties have in factadhered to the terms. It might seem that unforeseenevents would be problematic in this regard, because whatis appropriate behavior may not be obvious. The partiesmight then individually choose to act in ways that each


203

perceives to be perfectly appropriate but that the otherfinds non-cooperative and objectionable. This ambiguityshould be a real problem, however, only if the parties can-not consult one another and negotiate successfully what isto be done. While these negotiations might well be difficult,a bigger problem arises with informational asymmetries,whether of the hidden action or hidden knowledge type.

If the parties’ actions are not directly observable to oneanother, then it is possible that misbehavior will not bedetected or that it will be inferred when in fact it did nottake place. For example, suppose realized quality is notentirely under the control of the supplier. Then it may betempted to cut back on its quality control in hopes of notbeing caught, or, if caught, of managing to make the poorquality seem accidental. Moreover, even if the supplier isscrupulous about attempting to achieve the desired qual-ity, accidental failures may look to the purchasing partnerlike evidence that the supplier has cheated. In terms ofthe hidden information problem, if the costs and benefitsto each party from various courses of action are not rea-sonably clear to the other party, then what is in fact theoptimal thing to do is also unclear. This, too, can makecooperation hard to achieve. For example, if the supplier’scosts vary over time in ways the buyer cannot confirm, thenrequests for price adjustments can lead to very contentiousbargaining because it is unclear if they are justified.

The extensive information sharing that marks Toyota’sdealings with its suppliers helps minimize these asymme-tries and the resulting problems. Even after the relation-ship is well established, Toyota purchasing departmentengineers are frequent visitors to the supplier plants, andsuppliers are often at the Toyota facilities. Toyota has


204

extensive knowledge of the production processes at itssuppliers and, correspondingly, of their capabilities andcosts. Meanwhile, it shares its production schedulingplans with the suppliers and keeps these updated.

If there is a weakness in the system of supplier rela-tionships that Toyota has created, it is that Toyota may bein too strong a position relative to the suppliers. They arevery dependent on Toyota, especially in Japan, wheresuppliers to Toyota rarely also supply the other leadingautomakers. It would seem that Toyota could, if it chose,squeeze its suppliers, demanding unreasonable conces-sions and expropriating the returns they expected whenthey made investments to serve it. A supplier that hadbeen mistreated like this would not have a lot of recourse,and so such misbehavior on Toyota’s part might seemtempting. Then the threat of mistreatment—not to men-tion the actual experience of it—might lead the suppliersto behave less cooperatively or to expend resources tryingto protect themselves, either of which is value destroying.Yet, it does not seem to occur.

Suppliers want to work with Toyota—in an eleven-yearperiod, only three of the 176 members of kyohokai, the asso-ciation of Toyota’s Japanese suppliers, left the group(Asanuma 1989), while other firms (including GeneralMotors’ parts operation, now spun off as Delphi) have beenanxious to become Toyota suppliers. Moreover, Toyota’ssuppliers in Japan regularly make heavy investments thatare completely specific to Toyota, apparently undeterred byfear of hold-up. They typically do the actual design work tocreate products to Toyota’s specifications, and they own thedies that are used to make products to Toyota’s specificneeds. (In contrast, the norm in the American industry,


205

where short-term contractual relationships with outsidesuppliers prevailed until recently, was that the auto com-pany did any specialized design work, and if an outside sup-plier needed special dies to make a product, the autocompany would own these. These practices eliminated thethreat of hold-up of the supplier.) As well, there is econo-metric evidence that, as is asserted by the industry, Toyotaand the other Japanese auto companies actually do sharereturns with their suppliers (Kawasaki and McMillan 1987;Asanuma and Kikutani 1991). Finally, the fact that Toyotapurchasing engineers are welcomed into the supplier plantsand supplier personnel are often in the Toyota facilities indi-cates a fundamental trust.

Presumably, Toyota is sufficiently concerned with itsreputation that it does not act opportunistically. One fac-tor that may support this is that the suppliers are in con-tact with one another—indeed, Toyota itself encouragedthe leading suppliers to its Georgtown, Kentucky, plantto form a formal association. This effectively means thatany mistreatment of a single supplier will be known to all.If Toyota’s taking unfair advantage of one supplier wouldlead all of them to withhold full cooperation, this wouldsignificantly temper any incentives Toyota might havefor opportunism relative to what they would be if anypunishments were from just the individual firm it hadexploited. Further, the low interest rates at which Toyotahas been able to borrow mean that future returns are notlikely discounted heavily, so its reputation with partnerslooms large. This fits, of course, with the general long-term orientation for which the firm is known.

The importance of future returns’ not being heavily dis-counted means that maintaining a cooperative relationship


206

is easier when interest rates and, more generally, the costof capital are low. Thus, two weak firms are unlikely to cre-ate a strong, productive alliance, even if their capabilitiesand resources are very complementary. Future returns arenot sufficiently salient to offset current temptations andinduce full cooperation. And when a previously strongfirm gets in trouble, its partners need to be very wary.A partner facing a cash crisis is apt to behave in a veryshort-run oriented way, even if it would be completelytrustworthy in more normal times. When immediatesurvival is at stake, longer-term considerations are just notvery salient. Two examples from the auto industry illus-trate the difficulties.

General Motors in the late 1980s and early 1990s hadmade some moves towards establishing the sort of long-term relations with a select group of suppliers that we havebeen describing. It sought to engage them, to get them toshare cost information that they would normally have keptcompletely confidential, and to work cooperatively to cre-ate value. Then GM got in financial trouble—it ultimatelyregistered a loss of $23 billion in 1992. Part of the responsewas to use the cost information gained from suppliers toforce large price cuts. This breach of trust soured relationswith the suppliers for years afterward. Yet, it was com-pletely rational in the context of the crisis: Some GM exec-utives credited the money squeezed from the supplierswith saving the company from bankruptcy.

In such circumstances, it probably was in the interests ofthe many suppliers to help General Motors, voluntarilyadjusting the terms of the bargain to give the troubledpartner a temporarily bigger share. This would not onlyhave helped GM through its difficulties, it would also have


207

improved its incentives and might have helped maintaincooperation. The problem would have been to coordinateall this and overcome the free-rider problem among thesuppliers that each would like the others to help GM whileit kept to the original terms of the deal. This may be lessof an issue when the numbers are smaller.

The second example illustrates the possibility of offer-ing to help a partner in trouble, but it also shows the dan-gers of being dependent on such a partner (Burt 2002).The case involves the hold-up of Ford’s Land Rover sub-sidiary. UPF-Thompson, a small British firm that was thesole supplier of the chassis for Land Rover’s Discoverymodel, went into receivership in December 2001.KPMG, the accounting firm, was appointed to run thecompany while exploring a possible sale. Anxious to keepits supplier operating, Ford offered to increase the price itpaid for the chassis by 20 percent and to make an addi-tional “goodwill payment” of £4 million. KPMGrejected this offer, demanding that Ford make an up-frontpayment to UPF of £35 million and also increase theprice by substantially more than Ford had offered. Thetotal demand was estimated to amount to £61 million,although Land Rover had been spending only £16 mil-lion a year on the Discovery’s chassis. The KPMG part-ner in charge stated that his firm was merely fulfilling itslegal obligation to get the most from the assets of UPF:“Land Rover’s reliance on the company is an asset and weneed to obtain the best value on the asset.” Ford officialswere furious, and before obtaining an injunction requir-ing UPF to continue delivery, Ford was reported to havebeen considering ceasing production of the Discoveryrather than submit to the hold-up attempt.6


208

Even when there is nothing as dramatic as a bankruptcyto disturb arrangements, long-term relationships generallyneed to adjust over time. It is not sufficient to set the orig-inal terms of the understanding and then let it run, appeal-ing to the original contract when issues arise. At aminimum, as volumes depart from those that were forecast,prices may need to be adjusted to ensure an equitable divi-sion of the returns and thus maintain both parties’ interestin the relationship. Moreover, as technology, markets, com-petitors, customers, and the partners themselves change,the opportunities for cooperation will change, as will theparties’ outside options. All these affect the extent of coop-eration that can be maintained in the relationship, the divi-sions of the returns that are acceptable, and even theappropriate mix of activities for the parties to undertake. Itthus becomes crucial to the survival and success of therelationship that the parties be willing to adjust its terms.

The complex relationship among Xerox, Fuji PhotoFilm, and Fuji Xerox, a joint venture of the first twocompanies, provides an outstanding example of suchadaptation. Fuji Xerox was established in 1962 and con-tinues successfully today, one of the longest-survivinginternational partnerships or alliances (Gomes-Casseresand McQuade 1991). The roles of the parties, the activi-ties undertaken by each of them, the ownership structureof the joint venture, the nature and direction of paymentsbetween the parties, even the identity of the parties to thepartnership have all changed over the years in a flexible,adaptive fashion that has been crucial to the success ofthe arrangement.

Fuji Xerox was originally intended simply as a vehicleto sell Xerox’s revolutionary plain paper copiers in Japan.


209

It was in fact a creation of Rank Xerox, itself a joint ventureof Xerox and a British firm that had been set up to mar-ket Xerox products outside the United States at a timewhen the fledgling Xerox Corporation lacked resources toexploit its technology worldwide. A Japanese partner wassought to facilitate entry into the Japanese market, andFuji Photo Film was chosen, the only firm among thoseinterested that was not in the electronics business. Xeroxwas to provide the product, exported from America, andFuji Photo would staff the sales organization and providemarket knowledge. Government regulations mandated,however, that the joint venture had to do manufacturingas well as sales. Thus, Fuji Xerox was given a role in manu-facturing (largely assembly of knock-down kits fromAmerica), which it then outsourced to its parent, FujiPhoto.

The initial formal contract between Rank Xerox andFuji Photo Film specified payments from Fuji Xerox toXerox for use of the latter’s technology. It also providedfor protection of Xerox’s intellectual property—FujiPhoto Film had no right to use any Xerox technologyabout which it might learn through the joint venture. Infact, Xerox insisted on adhering strictly to this condition,refusing at one point even to consider ideas that FujiPhoto Film developed for new uses of the Xerox technol-ogy. At this point, Fuji Photo Film effectively became lit-tle more than a financial partner in the arrangement,passing manufacturing back to Fuji Xerox, but not with-drawing its money from the partnership or the people ithad sent.

This realignment of Fuji Photo Film’s role was justthe first of many changes in the basic structure of the


210

partnership. A few years later Xerox increased its owner-ship share in Rank Xerox to greater than 50 percent. Asa consequence, Fuji Xerox came under direct controlfrom Xerox headquarters rather than through the inter-mediary of Rank Xerox. In the early 1990s, Xerox andFuji Xerox set up a further joint venture, XeroxInternational Partners, to market Fuji Xerox’s laser print-ers in the United States. As well, responsibility for sellingXerox products in the Asia–Pacific region was shiftedfrom Rank Xerox to Fuji Xerox. In the late 1990s, Xeroxbought full ownership of the Rank Xerox joint venture, soFuji Xerox was now a 50–50 joint venture of Xerox andFuji Photo Film. Most recently, with Xerox facing finan-cial difficulties, Fuji Photo Film bought half of Xerox’sinterest in the joint venture.

That Fuji Xerox had laser printers to sell in the UnitedStates was a reflection of other fundamental changes thathad occurred in the relationship over time. From the out-set, Xerox had always been focused on developing andselling larger, faster copying machines aimed at corpora-tions that used them in central copying facilities. The dif-ficulties of reproducing documents written in Japanesekanji meant, however, that a larger part of the demand inthe Japanese market was for smaller machines that couldbe put in every office. When Xerox refused to developsuch machines, Fuji Xerox found money in its budget todevelop them itself, in direct contravention of Xeroxpolicy. The Fuji Xerox machines proved to be of highquality and sold well, and eventually they were acceptedby Xerox. In fact, Fuji Xerox came to provide all thelower volume copiers sold by the Xerox group through-out the world. This evolving role for Fuji Xerox was


211

recognized by repeated renegotiation of the contractualpayments between Xerox and Fuji Xerox.

Perhaps the greatest adjustment occurred in the flow ofknowledge between Xerox and Fuji Xerox. Xerox wonacclaim in the early 1990s as the first American companyto have gained back market share after losing a dominantposition to Japanese competitors. Fuji Xerox played ahuge role in this. Initially most learning flowed from theXerox’s technology to Fuji Xerox. Once Xerox lost theprotection of its key patents in an antitrust action, how-ever, it faced a wave of entry. The parent companyfocused on the threat from such established Americangiants as Eastman Kodak and IBM, but the Japaneseentrants to the industry were Fuji Xerox’s concern. TheJapanese in fact proved the greater threat to the Xeroxgroup as they came to dominate the lower end of the U.S.market and then to start making inroads into higher-volume copier segments. Fuji Xerox had already learnedthe techniques of designing and manufacturing productsat low cost and with exceptionally high quality thatbecame the competitive advantage of Japanese manufac-turers, including those like Canon and Ricoh that enteredthe copier business. Once Xerox finally awoke to the truenature of the competitive threat (after what it labeleda “lost decade”), it was able to import Fuji Xerox meth-ods of management and production, as well as Fuji Xeroxproducts, to reverse its decline. Thus, the parent learnedfrom the organizational child, as the flow of expertisereversed from its original direction.

These adjustments, involving the explicit contractualterms and the implicit relational contract, were key to thesurvival and success of the joint venture.


212

While evolution of any long-term relationship isinevitable, the parties need to careful about taking actionsthat would force adjustment in the partnership, becausethese can be destabilizing even when their intent was notto force changes in the terms. Choices that either partnermakes that alter the returns that the partnership can gen-erate can affect the possibility for continued cooperation.They could also undercut cooperation by causing thepartner to become concerned about its ability to protectitself against misbehavior. For example, developing anin-house capability in an area that has been the partner’sresponsibility reduces the difference in value between thewhat the partners can achieve by cooperation and whatthe firm can achieve on its own, making cooperation lessvaluable. It also may make the partner worry that it willbe replaced, which may deter it from committing futureresources to the partnership.

In a related vein, making the parties more dependenton one another or increasing the difficulty of severing therelationship generally tends to make cooperation easier toachieve. It was this effect that led telecommunicationsgiants AT&T and BT to avoid specifying up-front theterms for dissolution of their Concert joint venture,which was to serve the telecommunications needs of thetwo companies’ multinational business clients. But ifthere is some danger that external events, like regulatorychanges or demand shifts, could eliminate the basis forcooperation, then there is a substantial cost to such mea-sures. When demand for Concert’s services proved lessthan anticipated and the difficulties of integrating the twoparents’ businesses more substantial, the lack of agreedterms for dissolution proved very costly.


213

Despite such difficulties, the innovation of managingsupply through long-term relations has spread and playsa major role in explaining the increased outsourcing ofthe last decade.

Horizontal Scope

In the 1960s and 1970s, conglomerates—companiesinvolved in multiple, unrelated lines of business—were infashion. In the 1980s they fell from favor, and much ofthe merger and acquisitions activity of that decade in theUnited States essentially involved unwinding complexcombinations of distinct businesses created in the preced-ing decades.

The stock market appears to have believed in conglom-erates in the 1960s and 1970s, when they were first puttogether, for unrelated acquisitions were rewarded byincreases in market valuations that made the combinedentity more valuable than the pieces going into it(Matsusaka 1993). Later, demergers were also rewarded(Comment and Jarrell 1995), and scholars identifieda “conglomerate discount” (Montgomery and Wernerfeldt1988; Lang and Stultz 1994), with the market apparentlyvaluing firms in multiple lines of business at less than thesum of the values the component businesses would haveas stand-alone operations. While the extent to which themarket actually does apply such a discount is now a topicof some debate, it is clear that focus is much more in favorthan it once was. It would seem that even the wave ofmerger and acquisition activity that occurred in the late1990s did not lead to firms having as broad scopes as


214

twenty years ago, at least in the major English-speakingeconomies. Meanwhile, there are some signs that conti-nental European firms are starting to reduce their scope.Novartis and ICI are prime examples: Both have disposedof lines of business and become more focused. Earlier,Daimler Benz had done the same.

Why should putting two distinct businesses undercommon ownership and management affect the valuethey create? Or, equally, why should breaking them upinto separate businesses affect their value? What is theimpact of corporate strategy?

In an earlier era, the claim was made that combiningbusinesses under common ownership served a risk-reduction purpose: The diversified firm was like a diver-sified portfolio, with the aggregate being less risky thanany one of the pieces. Of course, this diversification is ofno particular value to shareholders, since they could holdboth firms’ shares on their own and achieve the sameresult. In fact, since the combination forces all sharehold-ers to hold the claims on the two earnings streams in fixedproportions, the merged entity actually restricts share-holders’ portfolio choices (although prominent asset pric-ing theories would suggest this latter factor may beunimportant).

Note that there are at least two points that mightcounter this powerful argument against the idea thatdiversification creates value by reducing risk. First, ifshareholders cannot easily diversify on their own, thenthere may be a risk-reduction value to having diversifiedfirms. This may, in fact, rationalize diversification in somecontexts. An example is family-owned firms, especially incountries with poorly developed capital markets. The


215

family cannot easily diversify through the financial marketsand does so instead by entering numerous lines of busi-ness. Second, while stockholders can diversify theirfinancial portfolios, the managers and employees of thefirm are not able to diversify their employment. If diver-sification allows for cross-subsidization of businessesdoing badly by ones that are doing well, and if this in turnreduces the likelihood of pay cuts and layoffs, then therisk faced by the employees is reduced. This has value,some of which might accrue to the shareholders if thelessened risk allows reducing average total compensation.

A diversified firm might also have advantages if it wereable to allocate resources between businesses more effec-tively than the market can. Williamson (1975) suggestedthat capital allocation might be done better within firmsthan across firm boundaries. Possibly the informationasymmetries that confront the internal capital allocationprocess are less troublesome than those in market deal-ings. For example, the manager of each unit might pri-vately know something about the value of extra capitalallocated to it. If the managers of different businesses arein the same firm, it may be easier to align their interestswith efficient allocation between the businesses thanwould be possible in a market. Then each may be moreinclined to reveal the relevant information, therebypermitting better capital allocation.

Similarly, if sensitive or subjective information aboutpeople and jobs flows more easily within a firm thanacross firm boundaries, it may be possible that the alloca-tion of human capital is more efficient within the firm.Even though, in comparison to the market, having fewerpeople to match with fewer jobs reduces the possibilities


216

for a great fit between the person and the job, the qualityof the average match may be better because of theimproved information. Certainly there is some reason toexpect that there are such advantages in communicationwithin the firm, because shared experiences and sharedlanguage make communication more effective. A diversi-fied firm may also be better positioned than the market tooffer managers a rich set of experiences in different busi-nesses, and thus may have some advantages in developinghuman capital. This argument has been made in the caseof General Electric, the most successful of the survivingconglomerates.

Diversification might also make sense if the firm hascapabilities or non-marketed resources that are not fullybeing utilized in its current line of business, yet furtherexpansion of that business is not attractive. An argumentmade in the 1960s by Harold Geneen, head of ITT, per-haps the archetypal conglomerate of the day, was that hisfirm had access to superior executives who together couldmanage assets and businesses better than others could.ITT’s expansion was then justified by the possibility ofletting its superior team of managers work with a largerset of resources to create greater value. Expansion in anysingle line of business was presumably restricted by thestrict antitrust enforcement policies of the day, and soITT became a conglomerate.7 More recently, Virgin’sexpansion of its scope, from airlines and trains throughmusic and bridal stores to soft drinks, may possibly berationalized by regarding the Virgin brand as such anincompletely utilized resource.

In a related vein, externalities among businessesmight also favor integrating the businesses inside a single


217

firm. Absent efficient Coasian bargaining or completecontracting, coordinating across businesses to handle theexternalities will require cooperation in the sense usedin Chapter 3—behavior that supports others’ welfare.Inducing cooperation may be easier inside the firm thanbetween separate organizations because the firm can offermore muted but better balanced incentives than thosethat come from dealings between distinct, autonomousparties. Thus, for example, if customer lists can be sharedor if a common brand gainfully used by different busi-nesses, or, more generally, if investments will affect mul-tiple businesses, then integration may have advantages.The Walt Disney Company’s expansion from its base inanimated films to television shows, theme parks, retailstores, cruise lines, and more, can be seen in this light. Ineach case, the characters and general brand characteris-tics (wholesome, family-oriented fun) were leveraged togain advantage in the new activity. This was more easilyaccomplished within a single firm because the brandcreated externalities among the businesses and theseneeded to be coordinated.

Complementarities can be a basis for interdependencebetween potentially separate businesses that can favortheir integration. The rationale offered by SonyCorporation for its acquisition of Columbia Pictures andCBS Records between 1987 and 1989 was based on thecomplementarity between “software” or content—films,TV shows, recorded music—and “hardware”—Sony’sconsumer electronics products that turned content intoentertainment (Avery, Roberts, and Zemsky 1993).Managing a company combining consumer electronicsand recorded entertainment has been problematic for


218

Sony. However, at the time of the acquisitions Sony hadhad a string of experiences that convinced its executivesthat the key to success for new consumer electronicsproducts might well be the ability to provide content thatcould be played on them. The film and recording compa-nies had historically resisted every new delivery technol-ogy that had emerged: Television would destroy the moviebusiness; audio cassettes would destroy the recordbusiness by permitting copying; home video cassetterecorders would destroy television and the movies, againby permitting copying; the compact disk would just beanother costly medium that would bring no extra demandfor recorded music. In fact, each innovation eventuallycreated huge new profits for the content providers.Meanwhile, Sony perceived that the CD had succeededonly because Sony and Philips, the co-developers of thetechnology, each controlled recording companies andrequired them to issue recordings in the new format.Further, some within Sony believed that Sony’s Betamaxmight have won the standards war for videocassetterecorders against JVC’s VHS standard if Sony had con-trolled recorded video entertainment that it could haveissued only in its format. Anticipating that it would con-tinue to develop new delivery systems for entertainment(including the then much-heralded High Definition TV),Sony bought control of the two content providers, hop-ing thereby to ensure the success of its future consumerelectronics products. Strikingly, Sony’s strategy was sooncopied by its archrival, Matsushita, which also bought astudio and record company.

The Sony experience points to some of the drawbacksof integrating diverse businesses within a single firm. In


219

fact, it appears that the hoped-for synergies neveremerged. None of the products Sony has developed in theintervening years seem to have benefited significantlyfrom its ownership of the entertainment businesses,although perhaps there has been some spillover betweenvideo games and the studio. There may yet be some valueto emerge from the combination if digitalization leads toconvergence of various industries including consumerelectronics and entertainment, as some now anticipate.Yet, the eventual returns would have to be immense tooffset the initial multibillion-dollar outlay and the subse-quent losses that Sony experienced and that led to a $3.4billion write-off on the acquisition.

The first problem was that the Sony leadership hadlittle experience with or understanding of the new busi-nesses. They knew something of the recorded music busi-ness from having had a joint venture with CBS Records inJapan, but Hollywood was unknown territory. Almost ofnecessity, they were forced to rely on managers hired intothe corporation from the entertainment industry. Theones they chose appear to have acted in a very opportunis-tic, self-interested way, assuring their Sony bosses that theextravagances they enjoyed and the failures they generatedwere the norm in the business. Perhaps the geographicdistance and the gaps in corporate and national cultures,compounded by choices Sony made to give the studioexecutives remarkable freedom and minimal oversight,made the difficulties uniquely severe in this case. Yet, thegeneral point remains that broadening the scope of afirm’s activities can make the top executives’ task of evalu-ating and controlling the individual businesses harder andinvites managerial moral hazard.


220

At the same time, Sony’s top executives were divertedfrom their previous single-minded focus on the consumerelectronics area. It may be only coincidence, but in theperiod after the acquisitions Sony experienced an unac-customed lack of success with a number of new prod-ucts—high definition television failed to take off, digitalaudio tape never caught on, the Minidisk® was a failure.Perhaps if the leadership had not been distracted by thenew businesses and if the amount of attention they couldgive to the traditional business had not been correspond-ingly limited, these failures might have been avoided ormitigated. Further, if the entertainment businesses hadnot been losing money there might have been moreresources to support investment in consumer electronics.

Having the disparate businesses in the same companyalso created tensions between the groups. For example,Sony’s head of consumer electronics in the United Statescomplained publicly about announcements affecting hisbusiness coming from the entertainment side. He eventu-ally left Sony, but that surely did not eliminate the problem.Top executives had to deal with these rivalries and con-flicts. Generally, heterogeneity invites corporate politicsand influence activities, with all the attendant costs.

Another problem comes in organizational design.Different businesses, with different technologies, mar-kets, and strategies, would on their own typically adoptdifferent organizational designs—different sorts of peo-ple, different architectures, different managerialprocesses and routines, and different cultures. Puttingthe businesses in a single firm means that either the firmmust deal with the complexity and difficulties of compar-ison that maintaining organizational differentiation


221

implies, or else the organizational designs must be mademore common, so that they are no longer optimized forthe individual businesses. Either choice is costly.

Sony opted to maintain as much differentiation as pos-sible. The result was complexity, increased influenceactivities, and loss of control. It is perhaps more commonto insist that standard operating procedures be appliedacross the company, precisely to avoid these problems andincrease comparability. Then the cost is likely to bediminished performance from having an organizationaldesign that does not fit any of the individual businesses’needs well.

A prominent example is Tenneco’s takeover of HoustonOil and Minerals (Williamson 1985: 158). Houston was inthe business of finding and developing hydrocarbondeposits. Tenneco was a conglomerate with some activi-ties in oil and gas exploration. Houston was known for thevery intense incentives it offered to its exploration per-sonnel, including giving them an interest in any petro-leum deposits they found. The spirit of the company wasvery entrepreneurial and its people were highly motivatedand very skilled, and Houston was extremely successfulat finding gas and oil. At the time of the acquisition,Tenneco vowed to run Houston separately from its otheroperations to maintain its entrepreneurial spirit, which ithoped would spread to the acquiring company. Withinmonths, however, Tenneco had imposed uniform corpo-rate processes and compensation systems on Houston. Anexecutive explained that common processes and proce-dures were imperative. The bulk of the talent that hadmade Houston so successful then left the firm. Eventuallywhat was left of the Houston unit was rolled up into the


222

existing operations. Other examples of the same sortabound.

Sony sought to limit influence activities and theproblem of complexity by its hands-off managementapproach. It did not assign any Japanese executives toHollywood and oversaw the entertainment businessesfrom an office in New York rather than from Tokyo head-quarters. This contributed, however, to the difficulties incontrol and in allocating resources effectively between thegroups.

Allocating capital between businesses is an especiallyimportant potential source of inefficiency in the diversi-fied firm. As noted, one of the early justifications of theconglomerates was the possibility that internal capitalallocation might actually be better than the market wouldachieve. More recently the billions of investment dollarslost by the dotcom start-ups and the collapse of the bub-ble in technology and telecommunications stocks haverevived the belief that the market may do a far-from-perfect job in allocating capital. So it seems possible thatinternal capital allocation processes might well do a betterjob. On the other hand, there are also reasons to suspectthat internal capital budgeting may be grossly inefficientand that separating businesses into distinct firms maythus create value.

The risk-reduction for managers and employees citedearlier as a possible advantage of multibusiness firmsmeans, a fortiori, that capital is not automatically beinginvested where it has its highest-value use. While theremay be gains in terms of the insurance offered to employ-ees, there is a cost in using capital in low-valued uses.Further, this cross-subsidization has an additional cost


223

that incentives are muted from what they would be in astand-alone operation. If losses in one unit are made upby resources generated in others, then the threat of bank-ruptcy and job loss becomes less motivating. Thus, theinternal allocation of capital becomes subject to a muchmore complex calculus.

Moreover, even if bankruptcy and job losses are not anissue, having multiple businesses in a single company mayresult in inefficient capital allocation because of influenceactivities or the desire to limit these. For example, in the1980s Nokia Corporation undertook a number of acqui-sitions to shift from its traditional focus on low-growthcommodity businesses (wood pulp, rubber goods, cables)to electronics. However, to maintain morale in the old-line businesses, they too were allowed to make acquisi-tions of their own. This expenditure came at a highcost—the debt Nokia took on for these acquisitionsnearly bankrupted it in the early 1990s—and the costswould certainly have been avoided if the old-line busi-nesses had not been part of the firm. (They were, in fact,shed after the crisis in the early 1990s, when Nokia choseto focus on telecommunications.)

The methods available to compete for capital differdepending on whether it is allocated by the market orthrough an administrative process. This means that theresulting outcomes may differ. In particular, the opportu-nities to deprecate or sabotage others’ projects are muchmore extensive inside organizations. It is clear that thissort of behavior can be costly in many ways.

An extensive empirical literature has developed on theefficiency or inefficiency of diversification. The earliestwork documented an apparent “diversification discount.”


224

The U.S. stock markets valued diversified firms at a lowermultiple of the replacement cost of their assets than morefocused firms (Montgomery and Wernerfeld 1988). Themarket also apparently valued diversified firms a loweramount than it would have given to the sum of their con-stituent parts (Lang and Stultz 1994). Later work repli-cated this last result, showed it also held in the UnitedKingdom and Japan, and traced it to inefficient internalcapital allocation processes (Berger and Ofek 1995; Shinand Stultz 1997; Scharfstein 1998; Lins and Servaes1999; Scharfstein and Stein 2000).

The technique used in many of these studies was tomatch each diversified firm with a collection of single-segment, stand-alone firms that, in aggregate, replicatedits range of businesses. Then the stock market perfor-mance and investment choices of the actual firms and theconstructed matches were compared. Doing so revealedthat the diversified firms apparently invested less instrong divisions with good prospects than did the focusedcompanies in these same lines of business and, corre-spondingly, overinvested in the weak ones. These distor-tions were typically traced to the impact of influenceactivities, which led to cross-subsidization of losers bywinners. Two possible mechanisms could have been atwork. The influence activities may have been successful intwisting choices, with businesses with limited opportuni-ties getting more resources than they should for efficiency(Meyer, Milgrom, and Roberts 1992; Rajan, Servaes, andZingales 2000; Scharfstein and Stein 2000). The otherpossibility is that the firms deliberately adopted evalua-tion and decision processes that distorted capital invest-ment in order to limit influence and reduce the attendant


225

costs, for example, dividing resources more evenly thanthey would for efficiency (Milgrom and Roberts 1990c).

The finding of a conglomerate discount has had amajor impact—management consultants refer to it inadvising their clients to “stick to their knitting,” limitingtheir companies’ scope to a narrow range of businesses.Many executives now believe there is a prima facie casethat diversification will reduce value, and there is indeedsubstantial evidence that the market rewards increases incorporate focus (Comment and Jarrell 1995; Berger andOfek 1996). Yet, much of U.S. employment remains indiversified firms, and almost as many firms increasedtheir diversification in the first half of the 1990s asreduced it (Villalonga 2002b). This presents two puz-zles—why would executives pursue apparently value-destroying activities, and how could they get away with it?

Very recently, empirical scholars have questioned thepresumed inefficiency of internal capital allocation andeven the existence of the diversification discount. Theproblem is that there may be biases in the processresearchers use in selecting groups of single-businessfirms against which to compare the investment choicesmade by the diversified firms and their market valuation.Measurement error may also be a factor, and togetherthese—rather than any real relative inefficiency or result-ing discount—may account for the statistical findings.

For example, Judith Chevalier (2002) examined theinvestment behavior of businesses in the 1990s that latermerged with one another. These firms showed the samepatterns of apparent over- and underinvestment whenindependent of each other as they did after combining.Thus, she concludes that the apparent distortions cannot


226

be attributable to cross-subsidization. Belen Villalonga(2002a, b), looking at the same period, found that theapparent distortions that were found when using standardmethods to select the single-business firms to match withthe diversified ones disappeared when she used moresophisticated methods. So, perhaps, there is no significantamount of inefficient cross-subsidization.

Further, Chevalier and Villalonga both found (as didCampa and Kedia (2002), who looked at a longer timeperiod) that diversifying acquisitions in the 1990s wereactually met with positive stock market reactions—themarket perceived them to be value-enhancing. This is inline with other studies that have consistently found posi-tive stock market returns to the acquired firm and essen-tially zero return to the acquiring one, implying that themarket perceives that value is created in aggregate in themergers. It also fits with a large number of studies thatsuggest that diversified firms are more profitable inaccounting terms. However, it is in some conflict withother studies that have shown somewhat negative returnsto diversifying combinations (offset by positive returns tocombinations that involve more closely related busi-nesses). Moreover, firms that diversified in the 1990sactually traded at a statistically significant discount beforethey diversified. Thus, if there is any causal linkage at allbetween poor performance and diversification, it appearsthat low performance leads to diversification, not theother way around.

Finally, there is the conundrum of General Electric—one of the most diversified of businesses and, throughoutthe last half of the twentieth century, consistently one ofthe best performing.


227

What then to make of all this? How can both acquisitionsthat increase diversification and those that increase focus begood? There is an obvious answer. If executives are actuallymaking choices that increase value, then changes in scope,whether they increase or decrease focus, should bring apositive response from the market. Some firms sometimescan increase value by exiting businesses, focusing on whatthey do best and reducing the costs of complexity andinfluence. Others can gain by increasing the range of busi-nesses, especially if the new business can take advantage ofunderutilized capabilities or resources or if they havesignificant externalities with existing businesses that cannoteffectively be handled contractually. When such moves aremade, value is also increased.

There has, however, been a strong suspicion amongmanagement scholars and economists that diversificationis often a value-destroying manifestation of moral hazardby executives. If executives enjoy heading bigger compa-nies and if the opportunities for growth in existing linesof business are limited, then the obvious choice is to growby increasing scope. This would seem to have been atwork in some cases, particularly in the oil industry in the1970s. High oil prices brought huge cash flows, whilenationalization of the oil industries in Africa and theMiddle East limited the multinational oil companies’opportunities to invest in the industry. Rather than givethe money back to shareholders, the oil companies diver-sified. Some of these investments were plausible, even ifthey proved unsuccessful. For example, several oil com-panies went into coal mining on the basis that it wasanother energy business based on extraction of subter-ranean hydrocarbons. Other moves seem completely


228

bizarre: One oil company bought a circus and anotherwent into fish farming!

Certainly there are cases where executive empire build-ing has occurred. Further, the executives pursuing it mayhave been behaving quite rationally from a personal pointof view. Leaders of bigger firms make more and havemore prestige, although it is unclear that acquiring otherfirms leads to higher pay. In any case, those who do acqui-sitions tend to get invited to serve on the boards of otherfirms (Avery, Chevalier, and Schaefer 1998). Until thescandals around corporate malfeasance erupted in2001–2, board memberships were plums. Yet, it seemsincorrect to assume that current diversification is primar-ily a matter of empire building. Indeed, if it were, thestock market should react negatively, and it does not.

The reason to reject a general presumption that diversi-fication is simply empire building is that, at least in theUnited States, executives of large firms now typically havesignificant stock and option holdings that should makethem very sensitive to the value of the firms they lead.Moreover, although changes in state laws have largelyremoved the threat of hostile takeovers that arguably drovemuch of the value-enhancing de-diversification in the1980s, corporate boards seem to have become much morediligent and demanding monitors of their executives, andsome institutional investors have become quite active inpushing for performance. So it would seem that executiveswould have little incentive or opportunity to indulge anytaste they might have for empire building if it is at anysignificant cost in performance.

Thus, we return to the logic that changes in scope arepresumptively aimed at creating greater value. In this


229

context, we would expect increased diversification to bepart of a strategy that emphasizes growth. Decreasedscope, on the other hand, would be consistent with a strat-egy that emphasizes increasing value by focusing on deliv-ering performance in current businesses. One might thenexpect to see increased focus being adopted in conjunctionwith other changes in organization that are designed toimprove current performance.

Further, if environmental changes allow new opportu-nities for growth in a firm’s core businesses, it might beexpected to focus on these and to leave other lines of busi-ness that it had entered when they were the only avenuesfor expansion. Thus, we might expect to see decreasedscope even among some firms that are focused on growth.Globalization is one such change: Lowering barriers totrade and investment and increased ease of communicat-ing, traveling, and shipping across borders mean thatcompanies have new opportunities to expand internation-ally and can grow by increasing their geography withoutincreasing the scope of products or services they offer.Deregulation and changes in antitrust policies have hadthe same effects in particular instances.

Internal Organization and Performance

While much economics research has addressed the verticaland horizontal boundaries of the firm and given a basis foranalyzing and evaluating the changes that we observe there,much less has dealt with the internal organization of thefirm and its impact on performance. Thus, our discussion


230

here must be much more speculative. Nevertheless, eco-nomic logic can shed some light on the sort of changes thatare occurring and their possible causes and effects.

The popular perception is that the technological andcompetitive environment of business is changing morerapidly than in the past, and business people assert thatmany of the changes they are making in their organiza-tions are responses to this acceleration. The logic theyoffer is similar to that we suggested in the case of theeighteenth-century Canadian fur trade. Recall that theNorthwest Company overcame the Hudson BayCompany’s huge cost advantage by creating an organiza-tion that empowered those close to the relevant informa-tion to make key strategic and operational decisions,motivated them via more intense financial incentives andownership to make the right decisions and to exertremarkable efforts, and created mechanisms to ensure thatinformation was shared and that the decisions of the manyindividuals were aligned. This worked spectacularly wellin the newly competitive environment and led to the nearcollapse of the HBC, with its more traditional, hierarchiccommand and control systems that had worked so well formore than a century while the HBC had been a monopoly.

The changes implemented at BP and in other looselycoupled, disaggregated organizations share many of thesame features. A number are particularly germane:

● establishing clarity about strategy and about corporatepolicies;

● creating discrete organizational units that are smallerthan previously favored;


231

● giving these units’ leaders increased operational andstrategic authority, and holding them strictly accountablefor results;

● reducing the number of layers in the hierarchy in aprocess of delayering;

● reducing the number of central staff positions;● increasing incentives for performance at the unit and

individual levels, perhaps accompanied by increasedrewards tied to overall performance;

● increasing the resources devoted to management train-ing and development;

● promoting horizontal linkages and communicationamong managers and staff, rather than requiring all com-munication to move up and down through the hierarchy;

● improving information systems that facilitate both themeasurement of performance and communicationacross units and up and down the hierarchy.

These moves are linked by a rich web of complemen-tary relationships, so the impact of adopting any one isincreased by doing the others as well. Thus, we tend tosee all being done together. Indeed, firms that adopt onlysome of them not only fail to achieve the substantial per-formance improvements experienced by those adoptingall; they may even suffer performance degradations.

Creating discrete, focused operating units and givingtheir leaders substantial decision rights over their activitiesshould have a number of direct effects that each improveperformance. The first is to improve the incentives for peo-ple in the unit to work hard on their unit’s performanceand think cleverly about their responsibilities—to showinitiative. This can come about in several ways.


232

First, knowing that you really have the responsibilityand authority to make decisions can be motivating initself. If the decision is surely yours to take, or even ifyour boss is just less likely to overrule you, you will caremore about the decision and probably invest more indeveloping the information needed to decide well(Aghion and Tirole 1997). Further, if hierarchic superi-ors have given up decision power, then your incentives tospend time and effort on influencing their decisions,rather than on getting your job done, are reduced(Milgrom and Roberts 1988b, 1990a). This can lead youto pay much more attention to your unit’s performance.

Delayering augments these incentive effects of creatingsmall, empowered units. Delayering typically means thatspans of control—the number of people reporting to anysingle manager—will increase, in particular for thosedirectly above any level that has been removed. This mayeffect a commitment by such managers not to intervene inthe lower-level decision-making, particularly if the num-ber of staff personnel has been cut, because they will nothave the time and resources. This, in turn, increases themotivating effect of reallocating decision rights.Moreover, if decision rights and authority are reallocatedaway from middle managers, there is less need for thesemanagers and so getting rid of them is more attractive.Thus, delayering and the creation of small, empoweredorganizational units are complementary in affectingincentives for initiative.

The second way that creating smaller units affectsmotivation is by making clearer the relationship betweenindividuals’ choices and actions and the performance oftheir unit. This can support both increased intrinsic


233

motivation and the provision of more intense formalincentives. If the organization is structured so that per-formance can be tracked and measured only for largesubunits, the link between what an individual does andthe results of these actions is very ambiguous. In a smallerunit, the linkage likely is clearer. This can increase intrin-sic motivation because you can see how your actions makea difference.

At the same time, measuring results for smaller unitsprobably improves the accuracy of performance mea-surement, reducing the randomness in any system linkingbehavior and rewards, since results are no longer buriedin an amalgam with the effects of other people’s andunits’ actions. This, in turn, allows giving more intenseformal incentives at the unit level without imposing toomuch risk. Thus, these features of the design are alsocomplementary, and creating empowered units makes itmore attractive to increase incentives. This is so both forexplicit financial rewards tied to performance and for lessformal rewards ranging from enhanced prospects ofpromotion to well directed praise.

Moreover, measuring performance for a smaller groupmakes providing explicit incentives to the group moreeffective by reducing the free-rider problem. This prob-lem arises because the returns to any improvements areeffectively shared among those whose performance ismeasured together (Holmström 1982b). When credit isshared less widely, the incentives are stronger. Finally,social norms to encourage performance-enhancingbehavior may also be more effective in smaller groups.

Increased incentive intensity, in turn, is complementarywith improving the measurement of performance: If


234

stronger incentives are indicated, then the benefits ofimproving performance measures increase. This can beachieved both by defining smaller, focused performanceunits and by improved information systems. At the sametime, as the quality of measurement increases, the attrac-tiveness of increasing incentive intensity goes up. Sofalling costs of measuring performance lead to moreintense incentives. Clarity about strategy, so that it is clearwhat constitutes good performance and what needs to bemeasured, thus leads to giving stronger incentives. Finally,improved measurement supports broadened spans of con-trol. Thus more of the common features are linked.

A second effect on performance of creating small,empowered units is to increase the likely speed of adapta-tion to new information, at least to the extent that theinformation is first available to those in the operatingunits rather than higher-ups in the organization.Information about customer needs, supplier and com-petitor behavior, and production conditions and opportu-nities should be most readily available to frontlinepersonnel, who are in direct contact with the sources ofthis information. (In contrast, information about emerg-ing political and regulatory issues, social trends, financialmarket conditions, and corporate policies is arguablymore likely to be available first at the corporate center.)Empowering those with the information to act upon itclearly speeds action, provided those with the decisionrights are motivated to make the decisions. There is noneed to wait while the information is communicated up,absorbed, and analyzed, and then the decisions sent backdown. Further, because aspects of the relevant informa-tion may be difficult to communicate and, in any case,


235

communication tends to result in garbling, the decisionsare likely to be based on more accurate information whenthey are made close to the source.

Increasing the speed of decisions is more valuable themore rapid the environment is changing. There is cer-tainly a popular perception among business people thatthe world is changing at an unprecedented and yet ever-accelerating rate. This then favors moving decision-mak-ing to the front lines.

One cost of empowering small, frontline units is thatquality of decision-making may worsen. In the traditionalhierarchic organization, a variety of decisions were madeby the middle managers whose jobs are eliminated indelayering. In particular, decisions affecting several unitsbut not the whole organization, such as the allocation ofcapital or customers between two divisions or whetherone unit can undertake a project that will harm another,would naturally have been made by an experienced man-ager who oversaw both units and who was responsible forthe results of both. Responsibility for such decisions nowmoves, either down to lower-level people or up to the top.Either can potentially be problematic.

If the decisions are moved down, the decision-makersmay not have the information needed to account forspillover effects on other units or the organization as awhole. They also may not be properly motivated toaccount for these effects. Further, to the extent that thenewly empowered frontline managers have less relevantexperience than the displaced middle managers, the qual-ity of decision-making may slip. On the other hand, if thedecision power moves up, then motivation is presumablynot a problem—top executives will want to advance the


236

performance of the overall organization. As well, since thetop management is no further from the front lines thanwere the now-gone middle managers, the amount andquality of information to which they can gain access is notlikely worse. This is especially true if information, mea-surement, and reporting systems have been strengthened,as is normally part of the set of organizational changes.Instead, the difficulty is that executive overload—toomuch to do and too many decisions to be made—may leadeither to hasty, bad decisions or to costly delays. There isalso a danger that the results of top-level interventionswill be applied on a blanket basis, undercutting the loosecoupling and the enhanced incentives that are importantelements of the model. This means that, in fact, the bulkof the decisions formerly made in the middle must movedown. Thus measures to improve decision-making byfront-line managers are complementary with delayering.

If the problem is that front-line managers do not under-stand the full implications of their decisions, one solutionis to provide the relevant information. This is increasinglyeasy with improved communication and information sys-tems, investment in which then makes the reallocation ofdecision authority more effective. Linking managers ofdifferent units directly so that they can inform oneanother about spillovers may be especially valuable, so thispractice is also complementary with the others in thepackage. Much of this may occur through informationtechnology, but face-to-face, personal contact is impor-tant, especially at the start when relationships are beingestablished. Being clear about strategy and about overallcorporate policy will also help, because it will set boundson decisions and will focus attention on key issues.


237

If the poor decisions come from misaligned incentives,the obvious answer is to provide the appropriate incen-tives. This might come from reducing the extent to whichperformance pay is linked to the individual unit perfor-mance and paying more on overall firm performance,although this may limit initiative and have negative impli-cations for unit performance. Establishing strong normswith regard to decision-making in the context of spilloverswill also be useful if they lead managers to account morefully for the effects of their decisions on others. BP’s tyingpay in part to overall corporate performance in part hadthis aim. Even more, its creating the peer groups workedto increase cooperation, because unit managers had towork with and rely upon the other managers who experi-enced many of the spillovers from their decisions.

On the other hand, Johnson & Johnson’s decentralizedstructure had great difficulties inducing cooperation amongthe thirteen J&J companies selling to hospitals in the 1980s.The company sought to introduce a shared service modelfor logistics and billing that would consolidate orders andshipments to the hospitals for the J&J companies. This wascompetitively important for many of the companies, whichwere losing market share, and would have generated impor-tant savings and increased convenience for the customers.The values of the company then should have induced coop-erative acceptance of the change, because these accentu-ated that the first duty was to the customer. However, thestrongest businesses with the best products, which werecompetitive even when the customers were not well served,were unwilling to give up any of their independence.

If limited capabilities are the source of the problem,then the first question is whether the center would do any


238

better. After all, it lacks the relevant information and mustdepend on the line managers for it. Even so, the solutionmay be in increasing the capability of the line throughinvestments in managerial development and the provisionof staff in the units. BP has, in fact, invested heavily inmanagement development.

A second difficulty that can arise with creating small,focused, empowered units is that the organization mayhave trouble responding to challenges and opportunitiesthat are most naturally addressed on some basis other thanthe “natural” linkage patterns derived from the primarydimensions of performance. At BPX, where the focus waslargely on increasing volumes and controlling costs, thelinkages among units connected ones facing similar tech-nical problems in achieving low costs and high volumes.Thus, units that were located in the same geographicalarea would not be linked together directly. But, for exam-ple, regulatory and environmental issues might be betterapproached on a national or regional basis.

The natural response to these problems is to creatematrix forms. These can vary in how explicit and formalthey are. ABB’s matrix of geography and product, withevery line manager reporting to two bosses, was at the oneextreme. General Motors employs a “basket-weave”model. The primary dimension of organization is byproduct and region, but senior executives take cross-unitresponsibility for processes such as quality and manufac-turing. BP tried having a Regions and Policies group thatwas nominally to be on a par with the business streams, butwould only have an assurance role. The business units,linked in their peer groups, would still report to the streamExecutive Committees, but the Regions and Policies staff


239

would be in contact with them on matters that did not fitwell into the business focus of the main structure. This wasnot fully successful, however. As the mergers and movesinto new business areas increased greatly the complexity ofthe company’s business and its geographic spread by thelate 1990s, it began to adjust its design to handle regionalissues better by giving additional, explicit regional respon-sibilities to senior managers overseeing the business units.

The Whole System

The complementarities described among the features ofinternal organization in fact extend to the horizontal andvertical scope of the firm. Clarity about strategy and themeaning of performance is easier when the firm is morefocused. A narrower scope also means that there is lessroom for the interdependencies among units that can beproblematic for the disaggregated model, although thelogic of partnering means that outsourcing does notrelieve the need to handle interdependencies effectively.Further, a small top executive team makes a narrowerscope more attractive, as the overload problem is less-ened. So all the pieces fit together.

The implications of these arguments have in fact beentested. Andrew Pettigrew and colleagues (Ruigrok et al.1999; Whittington et al. 1999) studied 448 European com-panies in the period 1992–6 to examine the linkagebetween their performance and the extent to which theyadopted elements of this model (the researchers focusedon ten elements involving changing the boundaries of thefirm, its internal architecture, and its processes). They


240

found a number of statistically significant positive correla-tions among the practices adopted in the best-performingcompanies in their set: Adopting one of the features (suchas greater horizontal linkages) was associated with adoptingothers (such as higher IT investments). Strikingly, theworst-performing companies often showed the reversepatterns, so that weak firms that had increased horizontallinkages were likely to be investing less in IT.

These observed patterns are in line with the argumentabove, with the existence of complementarities amongthe features of the design, and with the new model havingpositive performance effects. These results held up in afull econometric estimation of the determinants of per-formance. Adopting a coherent collection of the measureswas shown to have a strongly positive effect on perfor-mance, even though individual elements adopted on theirown often hurt performance.

The researchers also found that most companies in theirsample had begun to move to the new model, adopting atleast some of the collection of features. However, only atiny percentage (about one in twenty in the sample) hadgone all the way, adopting the full model by changing theirstructures, processes, and boundaries in fundamentalways. Those that had done so experienced much higherprofitability than firms that had not moved at all towardsthe new model. Strikingly, those firms that adopted onlyone or two of the three elements actually did substantiallyworse than those that had not moved towards it at all.

Thus, the changes that are being adopted, when seenand acted upon holistically, make excellent business sense.Together they improve performance substantially, whilemix and match does not work.


241

Notes

1. For a detailed examination of the range of changes and asurvey of the evidence that they are occurring, seeWhittington et al. (1999). See also Lichtenberg (1992) onfocus, McMillan (1995) on outsourcing, Rajan and Wulf(2002) for evidence on delayering, and Nagar (2002) on del-egation and incentive pay.

2. See Milgrom and Roberts (1992: 552–61) for a general dis-cussion of the issues involved. See also McMillan (1995)and Holmström and Roberts (1998).

3. Baker, Gibbons, and Murphy (2001), Levin (2003), andDoornik (2001) are recent exceptions.

4. The norm in the Japanese auto industry has been for onlysome 30 percent of the total cost of the vehicle to beincurred by the final manufacturer, with the rest being doneby outside suppliers. In contrast, General Motors in the1980s did 70 percent in-house, sourcing much less fromoutsiders. See McMillan (1995) and Milgrom and Roberts(1993) for more details and references.

5. The basic logic is very similar to that developed in the dis-cussion of reputations in the agency context in Chapter 4.

6. In fact, Ford had earlier stopped selling its Explorer sportsutility vehicle in Europe after Arthur Andersen, the receiverfor another insolvent British supplier, had demanded a60 percent price increase, with the threat to stop delivery ofcylinder heads for the Explorer. In that case the British HighCourt had ruled that Andersen had acted properly inexploiting the customer’s vulnerability, since the receiver’sresponsibility was to raise money to repay creditors.

7. Geneen also directly appealed to the risk-reductionargument.


242

243

6

Organizing for Growth andInnovation

Delivering current performance may please shareholders,but they are even happier if the firm is not just profitable,but if its profits and profitability are actually increasing.Moreover, the members of the firm are apt to like growthtoo. Work is more fun in a growing company. Growth meansmore opportunities for new, exciting assignments and pro-motions, and unpleasant conflict is lessened, because therecan be more for everyone from the ever-bigger pie that isbeing divided. Yet realizing growth without undercuttingcurrent performance and destroying value can be veryproblematic.

As long as the potential of the firm’s original businessis not exhausted, then delivering current performanceand achieving growth can be completely compatible. Thatis not to say that growth is necessarily easy. Size bringscomplexity, and organizational designs that were adequatein a small firm may fail to scale up successfully in a largerone. But there is no inherent conflict in extending thebasic strategic model as the business expands.

There are limits to such organic growth, however.Management scholars have argued that most industriesgo through a life cycle, from founding through a period ofrapid growth to maturity, when the industry essentiallygrows with the overall economy, and then perhaps toultimate decline.1 While no single firm necessarily followsthrough this whole cycle, because some enter late andothers exit before the industry dies, the dynamics of theindustry set the context for the growth of individualfirms. When the industry is growing fast, attracting newcustomers, there is lots of room for individual firms togrow rapidly too. Thus, some auto companies and theindustry as a whole grew very fast in the early twentiethcentury, while firms in the software, semiconductor, com-puter, and telecommunications industries grew attremendous rates in the later part of the century.

At some point, however, market growth must slow. Itmay still be possible for a few firms to keep growing fast byattracting customers from competitors, but this is likely tobe increasingly difficult. Another possibility is to expandinto new geographic markets, but there are often real lim-its to this approach. Leading a process of industry consol-idation by absorbing competitors can also bring continuedgrowth for some time. But there are obvious limits to thissource of growth as well: Even if further consolidationwould be economically possible, antitrust is likely to provea barrier.

Thus, if the firm is to continue to grow at substantialrates, it must ultimately be by developing business oppor-tunities beyond its original scope. Moreover, if demandfor a firm’s original products declines, then developingnew options can be crucial to its very survival.

Organizing for Growth

244

Acquisitions are one way to get into new businesses:Buy products, divisions, or companies already in thesefields. The other option is for the firm to create anddevelop new opportunities on its own. Acquisitions read-ily lead to higher sale numbers and higher total profits,but the implications for value creation and profitabilityare less clearly positive. Innovating, on the other hand, isan inherently uncertain process. Moreover, the organiza-tional designs that support it are very different from thosethat support delivery of current performance. This meansthat sustaining growth at rates exceeding the rate of growthof the economy is a difficult process—one that few firmsmanage successfully for very long. Meanwhile, hugeamounts of value may be destroyed in pursuing unattain-able growth ambitions.

This chapter is about organizing for growth. We willfirst discuss briefly the acquisition option. Then we turnto our main focus, the problem of organizing for innovationwhile continuing to deliver performance in the existingbusinesses.

Buying Growth

Acquisitions can certainly lead companies into new busi-nesses and new growth opportunities. General Electric hasundertaken hundreds of acquisitions. In recent years manyof these were by the company’s GE Capital financial ser-vices group, which in the process became a hugely diverseconcern in its own right whose growth was a major ele-ment of the parent’s. Acquisitions have also been used bysome companies to maintain the flow of new products


245

needed to compete and grow in high tech businesses. Mostprominently, Cisco Systems bought scores of smaller com-panies to obtain the technologies and people it needed tokeep innovating in its business in routers, switches andother network equipment, and software. Acquisitions haveeven been used to transform companies completely.Westinghouse Electric, the venerable American manufac-turing concern, in 1997 transmogrified itself into a mediacompany called CBS, owning the CBS television network,cable channels, and radio and television stations. It did thislargely by buying the media businesses and selling off itsoriginal manufacturing operations. Similarly, Mannesmannin 1990 epitomized German heavy industry, producingcoal, iron and steel, pipes, industrial machinery, and auto-motive equipment. Nine years later it was a pure-playmobile telecommunications provider. Beginning in 1990with a license to start a wireless communications network inGermany, Mannesmann grew to become a major player inthe industry by acquiring wireless companies in otherEuropean countries. In 1999 the company exited all itsindustrial businesses, completing its transformation.(Strikingly, both Mannesmann and CBS have since beenabsorbed by other companies, the former by Vodafone andthe latter by Viacom.)

The big problem with acquisition-based growth strate-gies is generating value for the acquiring company’sshareholders through the process. Profitless growth is notworthwhile, because just being bigger has no real value.In general, it seems very hard for firms to benefit theirowners by buying other firms. The difficulties are twofold.First is to avoid having all the potential value creationflow to the shareholders of the acquired firm. The second


246

is to realize the potential gains once the transaction iscompleted.

A consistent finding from a number of studies is that,when measured by stock-market reactions, mergers andacquisitions on average do create value, but most or all ofit accrues to the target firms’ owners. The general patternis that the stock price of the acquired firm shows signifi-cant positive “abnormal returns” (its price rises relativeto what would have been predicted from general marketmovements), reflecting investors’ perception that thetarget’s shareholders usually do very well, receiving a sig-nificant premium for their shares. On the other hand, thestock price of the acquiring firm tends to be unaffectedon average, or to fall somewhat.2 So, on net, investorsperceive that there is value created in the average deal, butthat little or none of it accrues to the acquiring firm.

This distribution of the surplus is what one wouldexpect if there is active competition to acquire companiesand potential targets are in short supply relative to thenumber of interested buyers. Of course, it is not inevitablethat sellers should be on the short side of the market, butit appears that they systematically are. Perhaps this isreflective of the desirability to executives of growing andkeeping one’s job over being acquired and possibly termi-nated. It may also reflect the targets’ actively seeking addi-tional suitors once the possibility of being acquired arises.Indeed, the directors of the target arguably have a duty totheir shareholders to try to get an auction going in whichthe price of their firm will be bid up by competition.

Even if competition for targets is not overly intense, aninformational problem may help explain the distribution ofthe benefits from acquisitions. It is known in the literature


247

on competitive bidding as “the winner’s curse.” The basicidea is that, in auctions where the value to the differentbidders of the thing being sold has a common element,unless the bidders are sophisticated there will be a tendencyfor the winner to end up paying more than the item is actu-ally worth. Target acquisitions may have this feature.

What a company is worth to any other company may beexpressed as some common amount that it is worth to anypossible acquirer, which reflects the underlying value ofthe business, plus an idiosyncratic component thatreflects the particularities of the fit between the target andthe individual bidder. No bidder likely knows either ofthese two elements with certainty. Instead, they must beestimated. If bidders with higher estimated values tend tobid more aggressively, then the winner will tend to be thefirm that estimates the sum to be largest. But even if eachfirm’s estimate of the value is unbiased, the largest of theestimates of the common component is systematicallylikely to be an overestimate of the actual amount. Unlessthe bidder takes account of the fact that, when it is thewinning bidder, it has the highest estimate of the value ofthe target (which means that it is likely an overestimate),it will tend to bid too aggressively and pay too much.

Sophisticated bidders may learn to avoid the winner’scurse by adjusting their bids downwards. For example,the oil companies have been aware of the problem formany years and presumably have learned to deal with it intheir bidding for oil leases. Whether companies biddingto acquire others have learned to overcome the winner’scurse is less certain.

Even if the winner’s curse is avoided, the fact of compe-tition for targets means that growth by acquisition can be


248

value-enhancing for the buyer only if the target is worthdiscernibly more to the acquirer than to others. In an auc-tion of the sort that often occurs in competing to acquirea company, the price will tend to reflect the highest valua-tion of the target among the losing bidders. Unless this isquite a bit lower than the winner’s valuation, all the gainswill flow to the acquired firm’s shareholders.

What makes a target worth more to one firm thananother? First, we need to distinguish whether we arethinking about the value to the executives of the acquireror the value to the shareholders. If the executives are build-ing empires, or if they are afflicted with hubris and overesti-mate their abilities to create value in running the target,these valuations may clearly differ. But as we argued in theprevious chapter, the rise of activist investors and theincreased reliance on executive compensation that is tied inone way or another to the stock price means that this formof managerial moral hazard is perhaps less of a problemthan it may have been in the past. Then the answer dependson two things: The potential for creating extra value byhaving the assets of the two companies under commondirection, and the actual ability to realize that extra value.

The potential for value creation in a business combina-tion depends on there being complementarities betweenthe two firms that cannot be realized except through uni-fied management and governance. In particular, if the gainscan be realized without combining the two firms—say, byan alliance, partnership, or contractual arrangement—thenputting the two firms together cannot be worthwhile.

The requisite complementarity could come from assetsthat are underutilized in one of the concerns and that canbe used in both. Many of the “synergies” that are often


249

claimed in mergers and acquisitions are of this sort:Having just one accounting or human resource depart-ment would allow savings if there are any scale economiesor indivisibilities and if the efficiencies these offer cannotbe realized by, for example, outsourcing these functions.Importantly, the relevant underused assets may be intan-gible ones, like knowledge. For example, it is arguable thatmuch of the value created in BP’s acquisitions of Amocoand Arco in the late 1990s came from applying BP’s supe-rior management systems to the American firms’ humanand physical assets. An attractive market position that canbe leveraged to include the new business might also be thesource of the gains. Cisco Systems could make moneybuying technology companies because of its unmatchedposition as a provider of network equipment to large busi-nesses and the reputation it had with its customers. Thesemeant it could get more from the technologies it boughtthan could another bidder. The complementarity gainsmay also come from each firm having assets that are mademore valuable by being used together. So if one has capa-bilities in manufacturing and another in marketing, thenpossibly the combined firm can do better than the sum ofthe parts.

The existence of complementarities that are economi-cally significant is plausible when the target and acquirerare in related businesses. For example, Newell Companyhas grown to be a major force in the home products busi-ness by buying other companies in the general field,including Rubbermaid, to expand its product line(Barnett and Reddy 1995). Newell adds value by applyingits unique management systems and by leveraging itsrelations with retailers. Such complementarities would


250

seem less likely, however, when the acquisition is meant totake the buyer into new businesses. Thus, achieving realvalue-creating growth in this fashion is apt to be difficult.

Moreover, even if there are potential complementar-ities in bringing two existing businesses together, actuallyrealizing them can be a major organizational challenge.Postmerger integration is notoriously difficult. It is easyto understand why.

There are basically three models for organizing after amerger or acquisition. One is to keep the two organiza-tions as separate as possible. This was the path initiallyfollowed by Sony after buying Columbia Pictures. Thisapproach largely gives up on realizing any significantcomplementarities, but it permits each part to have anorganizational design that fits its business. It may be app-ropriate if the objective is to get into a new business, per-haps to leave the old business behind as Mannesmann andWestinghouse/CBS did. However, it is hard to see howany value is going to be created in the process, and anypremium that was paid for the target will be difficult torecoup.

The other two approaches actually seek integration.One picks one of the two organization’s models (typicallythe acquirer’s) and attempts to institute it across theboard, with the target being reorganized and absorbed.This was done in the BP–Amoco merger, where the BPorganization design prevailed and the Amoco assets werebrought under the BP business unit model. The thirdapproach seeks to find the best in both organizationaldesigns and to combine them to create something quitedifferent. This is a common pattern, especially whencomparably sized firms are involved in a “merger of


251

equals.” The integration process following Hewlett-Packard’s acquisition of Compaq is a prominent example.

Imposing the organization of one of the firms has thegreat advantage of minimizing the confusion and disrup-tion that is involved in organizational change. It also con-tributes to rapid integration. There is a clear model, andthe new people simply have to adapt. Of course, theinclusion of a new group of people in the organizationwill necessarily change it, and their adoption of theselected model’s culture may be a slow and unsure process.This approach may work very well when the two compa-nies are in the same business, as in an industry consolida-tion like BP led. It is also the approach used by CiscoSystems in acquisitions of much smaller firms for tech-nology and human resources. Indeed, Cisco screenspotential targets with respect to the people and cultureelements of their organizations to ensure that they willeasily fit into Cisco (O’Reilly 1998). However, if theobjective is to bring in a really new set of business oppor-tunities, imposing the acquiring firm’s model is likely tobe very inefficient, because it is unlikely to be well adaptedto coordinating and motivating people to carry out theactivities of the new business. At the extreme, it risks driv-ing away the people who came with the acquisition, andthus possibly losing the capabilities that were the reasonfor the deal in the first place. The example of Tenneco andHouston Oil and Minerals cited earlier illustrates thisdanger.

The first danger in the “best of each” approach is thatchoices will be delayed and performance will consequentlysuffer for an extended period. What is the better wayneeds to be determined, and this may not be immediately


252

evident. Moreover, the choice can be an occasion forimmense politicking and resultant influence costs, as eachside competes to have the way it knows be chosen. Indeed,sometimes choices may never be made. Some twenty yearsafter the merger that created Japan’s Dai-ichi KangoBank, there were still separate Human Resource depart-ments in the bank to deal with the people who had comefrom each of the original companies. HP may actually havebenefited from the delay in carrying out its merger occa-sioned by shareholder opposition, because it gave extratime for the design issues to be settled before the actualcombination occurred.

The second problem is that, to the extent that the orig-inal designs involved coherent systems, the selection of theapparently best-performing features of each may not yielda coherent design. As we argued in Chapter 2, mix andmatch is often a recipe for disaster in the design problem.

The conclusion from all this is that trying to growprofitably through acquisitions is difficult, and that creat-ing value by getting into new businesses by this route maybe especially problematic. Thus, we turn to the problemof promoting innovation within established firms whilemaintaining performance in the original business.

Innovation in Established Firms: Exploringand Exploiting

Organizing to support the generation of new ideas is nothard. Research universities provide the model. Getbright, curious people together, give them time andresources and minimal direction, let them communicate


253

with other smart people who will both share thoughts andsubject ideas to rigorous examination, and make sure thatthe people whose ideas are judged best are rewarded in away they value (not necessarily with lots of money!). Theold Bell Laboratories of the American Telephone andTelegraph Company operated this way, and it generatedseveral Nobel prizes for the fundamental work donethere, including inventing the transistor and finding evi-dence of the Big Bang. Xerox’s Palo Alto Research Centerwas similarly designed and similarly productive: Object-oriented programming, local area networks, and manyof the key features of the personal computer were alldeveloped there. Of course, neither parent companymade any money from these great ideas—perhaps theywere organized a little too much like universities!

Thus, the problem is not just to create ideas, but also toturn them into successful businesses. Many establishedfirms struggle painfully to find new business opportuni-ties, and others never seriously try. Meanwhile, there areother firms that are constantly creating not just new prod-ucts, but whole new product categories. Yet, many ofthese firms that are so proficient at innovation struggle todeliver day-to-day performance. They can invent thingsand create novel businesses, but they seem unable to runthem with real efficiency. Firms that manage to be bothinnovative and efficient are rare. Our first concern will beto discover why this is so. We then discuss emergingresearch that suggests how to encourage the innovationneeded for growth without giving up too much in currentperformance.

James March (1991) has distinguished the two tasks thatare involved in firms creating new businesses and then


254

running them successfully. One is to exploit effectively theopportunities inherent in the current situation—the basicbusiness model the firm has adopted, the market segmentsit addresses, the products or services it offers, and thetechnology it employs. The other is to explore for anddevelop new opportunities. Exploration and exploitationare quite different tasks, calling on different organiza-tional capabilities and typically requiring differing organ-izational designs to effect them.

The essence of exploitation is achieving maximal per-formance in delivering on the current strategy. Asdescribed in the Chapter 5, consummate exploitationrequires complete focus on the current agenda, with allenergy aimed at effective, timely execution. Consequently,exploitation involves organizational designs—people,architecture, routines, and culture—that facilitate focusand execution. Incentives are likely to be strong and basedwhenever possible on quantitative measures of operationalperformance, costs, revenues, and profitability. Control ofprocesses is tight in order to reduce uncertainty and man-age risk. Meeting current customers’ recognized needs isaccentuated. Slack—resources not devoted to deliveringon the strategy—is ruthlessly eliminated.

Lincoln Electric has done a superb job of exploitation.Over many decades it has consistently and very effectivelypursued its aim of reducing the cost of the arc weldingequipment and supplies it makes, allowing it to reduceprices to customers and dominate its markets. Asdescribed earlier, everything in its organizational design isgeared to encouraging workers to increase the productiv-ity with which they carry out their assigned tasks. Thisruns from Lincoln’s famous piece rates and its remarkable


255

bonus pay scheme through to the way jobs are designedand on to the ownership structure. Every bit of cost isrelentlessly squeezed. Even what might appear to be wasteturns out to be a required element of the strategy for real-izing efficiency. For example, Lincoln’s unusually highlevels of work-in-process inventory are needed to allowindividual pacing of the work, which is crucial to the effec-tiveness of piece rates and the incentives they generate towork hard and be smart about how to do one’s job better.Similarly, the high level of employee earnings—they are atleast twice the average for manufacturing workers inLincoln’s home region—could not be reduced withoutundermining the trust and long-term loyalty of the work-force and without damaging the incentives for initiative(from the piece rates) and cooperation (from the bonus).

Exploitation is not just a matter of static efficiency in anarrow sense. Indeed, it can involve large amounts ofinnovation. Lincoln could not have achieved its record ofever-increasing output per hour without developing andimplementing innumerable process innovations over theyears. Similarly, BP Exploration, a superb exploiter, hasmade numerous major innovations in finding and extract-ing oil, including developing the ability to drill horizon-tally and in very deep ocean water. Moreover, firms intechnologically very dynamic industries have operated aseffective exploiters. For example, through the 1990s Intelexploited the opportunities in the X86/Pentium micro-processor architecture with remarkably single-mindeddetermination. Getting maximal yields in manufacturingfor each new generation of the chips was always a keypriority, so that costs could be minimized and price couldbe reduced. The development process for new chips was


256

highly regularized and disciplined. The focus on themicroprocessor business was complete—so much so thatthe desert creosote plant, which extrudes a poison thatkills any other plant that invades its space, became anaccepted metaphor describing the effect of the micro-processor business on any initiatives that might divertattention away from it (Burgelman 2002).

Although exploiters do innovate, radical innovations arenot the product of exploitation, but rather of exploration.Both exploitation and exploration involve searching forimprovement. But exploitative search is largely conductedin the course of normal business, looking for improve-ments in the context of the current agenda and model orfor fairly limited extensions of them. Exploration involvessearch over broader domains, looking for new opportuni-ties outside the current paradigm. It necessarily involvesmuch greater uncertainty, both as to whether anything willbe found and then whether it is actually better. It funda-mentally depends on slack—that resources be allocated touses that contribute little or nothing to executing on cur-rent strategy. Think of “New, Improved” Tide® detergentversus the digital compact disk. Proctor & Gamble’stinkering with one of their established products in a waythat intensive consumer research tells them will meetexpressed customer needs is search in an exploiter mode.Sony’s and Philips’s creating a new delivery system forentertainment whose advantages the potential customerscould not imagine is much more exploratory search.

A few companies have adopted a thoroughgoingexploratory orientation. The purest explorers focus com-pletely on generating ideas, leaving it to others to selectamong them as to which might be worth trying to develop


257

into businesses and then to create and run these busi-nesses. This is how research universities operate, derivingsome revenue from licencing their technologies, but fewcommercial organizations can afford this approach. Still,some come close. Rather than ever trying to build and runbusinesses themselves, they sell the ideas they generate orspin them off into freestanding companies. IdeaLab is anexample. Although IdeaLab has run into difficulties withthe collapse of the dotcom bubble, its original businessmodel was to create Internet businesses from the ideas ofits founder, Bill Gross, and company researchers. CEOsfor the new businesses were recruited from outside, andIdeaLab functioned largely in the now familiar role ofincubator, a role it helped create. It provided the initialidea, capital, office space, business services, specializedskills, and advice, and it later assisted in connecting toventure capitalists. The businesses became independentcompanies, with IdeaLab retaining a substantial owner-ship in each. As of 1998, there were over thirty indepen-dent companies that had come out of this process.

Few companies are content to be pure explorers orincubators, and yet they still focus heavily on exploration.One of the most notable is 3M, especially in the periodthrough the 1960s (Bartlett and Mohammed 1995). Thiscompany has a huge array of products, from the wet-usesandpaper it developed a century ago, through the reflec-tive signs used worldwide on freeways, to Scotch brandadhesive tape, and medical supplies. Its forte has beendeveloping new, breakthrough products. This requiresimagination, thinking “outside the box,” a willingness totake significant risks and to accept failures (and evencelebrate, rather than punish, ones that had a chance),


258

openness to the new and untried, and finally the slackresources to generate and develop ideas that may initiallyseem very far from successful product offerings. 3M puttogether its organization to make all this possible.

Multiple R&D groups were established, and there waslittle attempt to coordinate and rationalize activity acrossthem. Direct communication among groups was stronglyencouraged, however, and became the norm. Thus, athousand flowers were let bloom, and multiple approachesto a problem could co-exist, compete, and cross-fertilizeone another. 3M also gave its people unusual freedom andautonomy in deciding how and on what to work. To pro-tect this freedom, it actually mandated that technical peo-ple should be able to devote 15 percent of their time to“bootleg” projects of their own choice, rather thanworking on the ones to which they were officially assigned.The ubiquitous Post-It® note came out of such efforts.Moreover, those who resisted directives from senior man-agers to abandon projects and instead carried themthrough to success were heroes in the organization. Eventhe CEO told stories lauding such rebels who had directlydisobeyed him when he had tried to kill their pet projects.

At a corporate level, objectives were set for the share ofrevenues coming from new products. Performance mea-surement for subgroups or individuals, however, tended toinvolve subjective evaluations or milestones achieved, notthe financial numbers generated. Rewards had large non-monetary elements, especially personal autonomy andprofessional recognition, and a separate scientific andengineering career path was created so technical peoplecould advance without having to move into management.The sales people were charged as much with being a


259

communication link between customers and the labs as withselling products, and their pay was arranged accordingly.

On the more mundane side of business, the motto was“make a little, sell a little.” 3M did not attempt to imposethe sort of discipline that was needed to compete in massmarkets on the basis of cost. Rather, it pursued nichemarkets, abandoning these when competition from imita-tors proved too intense.

Both 3M and Lincoln are extreme types, and suchspecialization is rare. The reasons are clear. An exploreris very vulnerable to fast-follower competition that operatesmore efficiently and can undercut the originator of theproduct, stealing the market before the costs of develop-ment can be recouped. Indeed, 3M in the 1990s facedincreasingly tough competition of this form, leading it tospin off its videotape and computer diskette businesses.More generally, the 3M model generates lots of ideas, butdoes not rigorously select among them, and so costs areinflated by the resources sunk over extended periods onideas that will never become products. Meanwhile, long-run survival as a pure exploiter requires more than beingvery good at executing on the strategy. In addition, thestrategy must remain viable. This means that there cannotbe major shifts in demand and that superior new tech-nologies do not undercut the firm’s competitive advan-tage. This latter danger has threatened Lincoln in the lastdecade as new competitors have introduced weldingequipment using new, lighter materials and new elec-tronic controls. The competitors have also moved to supplycustomers on much shorter lead times than Lincoln canguarantee because its policy against layoffs constrains itshiring in busy periods. Thus, firms need to mix elements of


260

explorer and exploiter if they want to innovate and growprofitably on an ongoing basis.

Even in the simplest case—a company that comes intobeing to create, develop, and sell just a single product andthat is content to go out of business if and when the demanddries up—the firm must move between exploring andexploiting as it develops. The early stages of the firm’s liferequire it to operate in explorer mode, but then it must shiftto become an exploiter if it is to prosper. This in itself maybe problematic. Venture capitalists insist on having the rightto appoint the CEOs of their portfolio companies, and theymake frequent use of this power to replace companies’founders (Hellmann 1998). They do this presumably inpart because the characteristics needed to come up with theoriginal business idea and prove its value are not the sameas those required to develop the idea into a functioningbusiness or to run the business effectively once it is estab-lished. Organizationally there may be problems as well. Thepeople who were attracted to the firm in the freewheelingearly days of exploration may not fit in the disciplined,focused exploiter context. The communication patternsand decision processes that were appropriate in the smallstart-up may need to be replaced by more formalized,bureaucratic ones as the firm grows. The culture of work-ing together to create something must be replaced by one oftaking responsibility for executing assigned tasks. All thiscan be disruptive. However, once our single-product firmmakes it to the stage of exploitation, it can focus on that onemode of operation until it ultimately ceases to exist whenthe product cycle ends. As the Lincoln and Hudson BayCompany examples make clear, this can be a long time—Lincoln has been in the arc welding equipment business


261

since 1895, and the HBC was still buying furs and sellingmanufactured goods three centuries after its founding.

To move beyond this model, where growth prospects aretied so completely to the single product’s market, firmsmust develop multiple business opportunities, and to con-tinue to grow and survive they must do this on an ongoingbasis. Thus, at any time they must have a portfolio of activ-ities going on within the firm: Searching for new oppor-tunities, selecting from among identified opportunities,building new businesses, running existing ones, exitingothers—all may need to be done at once. In particular,they must explore and exploit at once. This puts the com-panies right in the middle of the multi-tasking problem ofexploring and exploiting simultaneously. Striking andmaintaining an appropriate balance between explorationand exploitation can be hard. Moreover, the better you areat one, the harder it is to do the other well.

The first difficulty with multi-tasking is the motiva-tional problem that was discussed in detail in Chapter 4.The difficulty is in inducing employees to allocate theirtime, effort, and attention appropriately among differenttasks when these differ significantly in the timeliness andaccuracy of the available performance measures, as is thecase with exploration and exploitation.

If two tasks compete with one another for employees’time and attention, the incentives offered for each must beof comparable intensity, or else the employees will tend tooveremphasize the well-compensated one and under-supply the other. In the extreme, the employees will justignore the less well-rewarded one completely, because thatis the way to earn the highest rewards from any given levelof total effort. Providing comparably intense incentives for


262

different activities becomes problematic, however, whenthe available measures of the two tasks differ greatly intheir precision or timeliness.

Generally, the better the performance measures on anactivity, the less costly it is to provide stronger incentives for the activity in isolation and so to induce higher effort. A major reason for this is that the poor measures mean theemployee faces large amounts of uncontrollable risk in herrewards, which will vary not just with her choices, but alsowith the randomness in the measures. This uncertainty iscostly to a risk-averse person, and the cost rises as the incen-tives become more intense, causing the amount of uncer-tainty in the rewards to scale up. Since the agent must becompensated for bearing this risk, good measures tend tolead to strong incentives and poor measures to weak ones.

When the quality of the measures on two activitiesdiffers significantly, it may thus be extremely difficult orexpensive to give strong incentives for the ill-measuredactivity, even though it is easy to provide intense incen-tives for the other. Thus, providing strong, balancedincentives for both can be problematic. Consequently, ifboth tasks are desired, it may thus be necessary to supplyrelatively weak—but balanced—incentives for both.(Indeed, theory suggests that the optimal incentives toinduce multi-tasking may be even weaker than those thatwould be given for the poorly measured activity in isola-tion.) The weakness of the incentives then means theagent will not devote much effort to either task, even thewell-measured one to which she could be induced todevote significant effort if it were in isolation.

Exploratory activity is typically hard to measure in anyprecise and timely way. The appropriate behaviors are


263

hard to specify in advance, the connection between theemployee’s efforts and the results that are achieved issubject to real randomness and may be poorly under-stood, and the actual value of the results may not becomeclear for a long time. Exploitation, on the other hand, ismuch more easily measured. Sometimes the tasks are sowell understood that the desired behavior can be speci-fied in advance and controlled directly. More generally,the connection between effort and results is likely to bemuch clearer and less noisy, and the measures of perfor-mance—operational and financial results—are readilyavailable. Moreover, the tough, disciplined mindsetappropriate to generating current performance is likelyrather different from the more open, experimental onethat might go with generating new ideas in previouslyunexplored fields.

Consequently, if current performance is desired, quitestrong incentives can be given and significant effortinduced at a reasonable cost. On the other hand, if onlyinnovation is desired, it too can be induced, althoughmaking the incentives very intense can be quite costly.Getting both sorts of activity from employees can be aproblem, however. This is especially true in a contextwhere one of the two modes of operation has been theprimary focus, and the company now wants to do both.

In many firms in the last decade, the focus has been oncost control, and appropriately intense incentives havebeen provided for this one activity. Suppose one of thesefirms decides it wants to encourage growth, and for this itsees that it needs to induce more innovation fromemployees. Yet, it is loath to give up the performance ithas been enjoying in cost control. What can it do?


264

Note first that simply adding some incentives for innova-tion to the employees’ existing reward package is unlikelyto have any effect. Until the incentives for attemptinginnovation are comparably intense to those for cost con-trol, the employees will rationally ignore them, becausediverting thought and effort towards innovation willlower their expected rewards. It may also increase thevariability of the rewards and thus the risk they face. Thismay help account for the often-heard complaint fromexecutives that, despite their being eager to welcomegrowth-promoting ideas, their people just seem to lackthe imagination to come up with anything. The fact is, forall the talk, the inducements offered just do not justifydiverting attention away from what is really rewardedtowards risky and poorly compensated activities.

As the incentives for innovation are increased, at somepoint they may become sufficiently intense that they doinduce the employees to allocate some attention in thisdirection. But because the two activities compete for theirtime, unless the incentives for cost control are alsoincreased, the effort and expected performance in thisactivity will fall. This is a second sense in which multi-tasking is problematic—inducing one sort of behaviorincreases the cost of getting the other. Moreover, thenow-intense incentives are tied to noisy, imperfect mea-sures. This means the employees’ rewards vary signifi-cantly with the randomness in the measured performanceand only partially with their own efforts. This variabilityis costly to risk-averse employees. The cost might beexpressed in terms of discomfort with the lack of controlthey experience. Their rewards vary immensely, but notwith their efforts! Then either expected total compensation


265

must be increased, which is directly costly to the firm, orelse the employees may quit.

Both these alternatives are actually taken in organiza-tions. Many start-ups in the 1990s used stock options to give very intense incentives for exploratory activity.Even given the probability that the options would proveworthless, the expected returns on them—and thus theexpected total compensation offered to employees—waslikely much higher than they could expect elsewhere (or, at least, the employees who chose to join such firmsperceived this to be the case). This high expected mone-tary value was needed to offset the risk that the employ-ees faced. The “solution” of losing people is adoptedimplicitly when firms do not give adequate returns to off-set the risks they attempt to load on employees when theyask them to undertake risky projects. The employees facethe risks to their incomes and career prospects thataccompany failure but get little of the returns that successgenerates. Many just refuse the offer, either by quittingor, more subtly, by diverting attention to lower risk activ-ities. In either case the firm in fact fails to get the innova-tion it claimed to want.

These factors help explain the problems that somefirms have experienced in trying to become more innova-tive. For example, despite very large investments in R&D,Proctor & Gamble, the U.S. consumer products firm,regularly failed through the 1990s to develop and suc-cessfully introduce really new, innovative products. Latein the decade it decided to reorganize to speed new prod-uct development and introduction. In the process, how-ever, it lost control of costs, revenues did not climb asquickly as they had (let alone as fast as had been hoped),


266

earnings fell, and the stock price collapsed. All this led tothe CEO’s replacement in June 2000, after less than twoyears on the job.

Related problems arise if an explorer tries to push forgreater operational efficiency. The natural way to do thisis to try to introduce a little more focus and discipline andto try to weed out some of the slack. For example, man-agement may start measuring operational performanceand talking about the need to control costs. Such movesmay have little effect, because the incentives are unbal-anced. This is obviously unfortunate. More serious is thepossibility that the measures will work, but that in theprocess they destroy innovation.

In an organization that is giving relatively intense incen-tives for exploration, as through stock options in a start-up,introducing limited incentives for current performancemay have little effect, because the main rewards are still forthe innovative activity. Moreover, it is likely that the peoplewho were attracted to the firm when it offered suchrewards will not have a great taste for the perceiveddrudgery of exploitation, and they will be inclined toignore calls for it, even if these come with some explicitrewards. After all, they came to the company to work oncool things and to get a chance at the big prize; respondingto calls for boring cost control is not what they want to do.

The situation is different in more established explorers,like 3M, where the incentives have probably been moremuted. People have likely been motivated primarily bypride in their work and the intrinsic pleasure of it, andthey likely value the freedom that they enjoy. The naturalmeans to increase efficiency, such as tighter controls,more disciplined resource allocation, and more explicit


267

rewards tied to outcomes, are likely to have a significanteffect. In particular, they may shift attention radically. Ineffect, the failures that must be common in reallyexploratory activity now become very costly to thoseinvolved. This can reduce risk-taking immensely andseverely curtail innovation. Moreover, the problem ofmismatch between what attracted the current people tothe firm and what it now offers them is likely to be at leastas intense as in the start-up.

The last several chief executives at 3M have sought toincrease efficiency while not destroying the innovativeengine that was their company. Thus, while they institutedmore stringent reviews of research projects and speededup dropping unpromising ones, they also increasedresearch budgets significantly. However, they were under-taking a tough balancing act, and there have been sugges-tions that the effects on innovation were negative.

Job Design for Multi-tasking

One obvious solution to the multi-tasking problem maybe to divide the jobs, with some employees exploring andsome exploiting. Then each has a simple agenda, and nei-ther faces any conflict between tasks. There may be draw-backs, however, if it would otherwise be advantageous tohave one person or group undertake both activities. Forexample, the probability of a successful innovation maydepend on knowledge of the current technology or mar-kets that can best be gained from working in the area.

Other motivational problems arise when a business unitseeks to pursue multiple goals and assigns different groups


268

within the unit to each activity. These problems have to dowith internal competition within the unit and influenceactivities. The problems are likely to be especially severe ifgrowth and innovation are added to the agenda of a unitthat has been focused on current performance.

If a single unit is expected both to push performance oncurrent products and services and to develop new ones,the teams assigned to each task are likely to be in compe-tition over resources. Funds can go into developing newproducts or making and marketing current ones. Talentedpeople can be assigned to one group or the other. Theattention of the unit leaders can be allocated to support-ing the growth team or the performance team. The nextpromotion can go to a member of one team or of theother. This internal competition is apt to be costly anddivisive under the best of circumstances. Each side will betempted not just to argue the merits of its own case, butalso to denigrate the other. Each may become morefocused on the internal competition than on the externalcompetitors and customers. Resources earned from theefforts of the team handling the current offerings arespent on the pet development projects of the other team,generating jealousy and discontent. When the productsbeing developed would themselves compete with the cur-rent offerings, the competition can become really intenseand destructive. The results can be inadequate attentionbeing given to either task, compounded with severemorale problems.

Moreover, it really is not possible to eliminate themulti-tasking problem by assigning exploration to onegroup and exploitation to another. If the firm is going toboth explore and exploit, someone will have to multi-task.


269

At a minimum, the head of the organizational unit who isresponsible for motivating its members to do the differenttasks must face multiple objectives. So even if each of thepeople in the unit is given a simple task, the head mustface balanced incentives if she is to be induced to motivatethe units’ people to undertake both sorts of activity in theappropriate amounts.

If the unit leader’s rewards could be structured so thatthey depended on the full returns generated by the unitover time, then from the point of view of inducing multi-tasking it would be equivalent to the leader’s making deci-sions on her own account. There would then be littleproblem in motivating the manager’s multi-tasking(although there might be trouble motivating those belowher). In reality, however, business unit managers rarelyface such incentives. Their rewards are typically tied tooverall corporate performance, to unit accountingreturns, and perhaps to some subjective measures or thepassing of various milestones. Tying results to corporateperformance, even if it is measured in stock prices, buriesthe impact on rewards of any results generated by theindividual unit itself and creates a free-rider problem.Accounting measures are a poor proxy for the value cre-ated, let alone for the effort and imagination that has beenemployed, especially in a context where innovation isbeing pursued. Indeed, current accounting profits arelikely to be hurt by innovative search. Subjective mea-sures and milestones may provide more effective incen-tives for innovation than do the accounting numbers, butusing them to provide very intense incentives is certainlyproblematic. Big rewards tied to subjective evaluations arean invitation to politicking and their administration can


270

easily seem capricious, biased, and unfair. Consequently,motivating unit managers to pursue a balanced agenda ofboth performance and growth is tough.

On the other hand, top executives may be appropriatelymotivated to make the right choices between explorationand exploitation. At least in the United States, typically alarge fraction of their substantial compensation is tied tothe firm’s stock price. To the extent that the stock pricerecognizes both the short- and long-run prospects of thefirm, executives do have strong, balanced incentives.Thus, it might seem that an easy solution to the multi-tasking problem is to have only the top executives respon-sible for both exploration and exploitation. Thenindividual business units and functions can focus onsimple agendas of performance or growth.

Typically, this would involve creating different units topursue projects outside the normal scope of the existingbusinesses.3 The existing units focus on exploitation ofcurrent opportunities, while the new units explore for newones and build them into businesses. Creating a separateexploratory unit also facilitates application of processesand measures that are appropriate to a new business andmay help ensure that it gets the needed managerial atten-tion. However, it does not eliminate the conflicts betweenthose charged with delivering performance in currentbusinesses and those seeking to build new ones. Instead,the battleground shifts to the corporate level, and theresultant influence costs may be even greater.

To protect the new business from the “creosote plant”effect of the existing ones, some companies try to separatethem radically via the “skunk works” model. The termreferred initially to teams working on top-secret defense


271

projects at Lockheed. To maintain secrecy, the teams wereisolated from the rest of the organization. The model hassince been adapted and widely adopted. IBM developedits first PC using a new ad hoc unit that was located farfrom other IBM facilities and whose existence and man-date were kept a secret from other parts of the firm. Theidea was to protect the project from the dominant cultureof the mainframe computer business, then IBM’s core.Similarly, General Motors created its Saturn Division tobe “a new kind of car company” producing “a new kindof car.” To protect it and allow it to develop new ways ofworking and of relating to suppliers and employees,Saturn was located far from GM’s Detroit home base.The question with the skunk works model is whether thenew unit can ever be integrated back in successfully. IBMsucceeded, but GM has struggled with Saturn. The newunit will still meet resistance and possibly jealousy, and tothe extent it is organized differently from the rest of thecompany, there is complexity and the opportunity forinfluence costs.

There are other difficulties, moreover, with passingresponsibility for multi-tasking to the top levels of thefirm. Most obviously, the top executives may be farremoved from having first-hand knowledge on which tobase allocation decisions. This is true under the best of cir-cumstances in large firms, and when the decisions involvesuch matters as the returns to different lines of explo-ration, it is even more likely to obtain. Thus, the decisionswill necessarily be based on second-hand knowledge andon hunches and gut-feel. This can obviously result in lessthan ideal decisions. Moreover, the information thatthe executives will have for decision-making will be


272

very asymmetric. Exploration will have a difficult timequantifying the costs and benefits that might flow fromgiving it resources, while the exploitation groups will beable to document their cases well. This may tend to biasdecisions in favor of the established, performance-oriented businesses.

To compensate, executives may try to adopt a biastowards the exploratory, but this too can cause problems.When a project gets the CEO’s early support, it will tendto get resources that it might not objectively deserve. Forexample, Intel spent hundreds of millions on a cameraproduct to permit video-conferencing using personalcomputers (Burgelman 2002). CEO Andy Grove’s beliefin the product kept resources flowing to it despite signsthat it would be a failure in the market (which it was).Similarly, Apple’s Newton personal digital assistant wasthe pet project of CEO John Sculley. Although the prod-uct was good in concept (as Palm later showed), the fail-ure of the Newton itself might have been forecast, andhuge amounts of resources saved, but for the CEO’s pro-moting it. Even absent such problems, the top executives’time is limited, and this may result in delays in decisions,overload, and frustration for all concerned.

An orthogonal approach to the problem of multi-tasking is to try to change the fundamental trade-offsinvolved, largely by working on the people and culturalelements of organizational design. The idea is to try tobring greater alignment between the interests of the firmand its employees so that the agency and incentive prob-lems that underlie the multi-tasking problem as we havediscussed it are lessened or even eliminated (Day et al.2002). The general term used for this approach is “high


273

commitment” human resource management. As arguedin Chapter 4, if only very weak explicit performancerewards are possible, then it may be especially desirable toincur the costs of adopting high commitment HRMbecause it may lead to higher levels of effort (and a betterallocation of it among tasks) than can be induced by theweak explicit rewards. The hallmarks of this system aretrust, transparency, empowerment, egalitarianism, jobenrichment, teamwork, the absence of explicit individualmonitoring and performance pay, and employees’ identi-fying their interests and those of the firm and acceptingits vision wholeheartedly.

Nokia Corporation has used many of these organiza-tional levers with some success in achieving both innova-tion and remarkable efficiency in operations. In 1992Nokia was a failing Finnish conglomerate whose productsranged from rubber boots through wood pulp to televi-sion sets. The collapse of the Soviet Union had sent theFinnish economy into severe depression, the rest ofEurope was experiencing a sharp recession, and both thecompany and the Finnish banks that were its leadingshareholders were in serious financial difficulties. In thiscrisis it offered to sell its small mobile phone business toEricsson, but the Swedish company was not interested.By 2000, Nokia had Europe’s highest market capitaliza-tion and the fifth most valuable brand in the world.Completely focused on telecommunications, it was theclear global leader in mobile phone sales and a strongplayer in providing the supporting network equipment.Its margins on phones were estimated to exceed 20 percent,while those of its chief rivals—Ericsson and Motorola—were at best in the low single digits. Nokia had achieved


274

this transformation by consistently leading the industryin developing new models with better technology,features, and design at the same time that it maintainedthe strict control over operations necessary to keep meet-ing the exploding demand for its products. Even after theboom in telecommunications collapsed and many firms inthe telecommunications equipment industry ran intosevere difficulties, Nokia continued to thrive.

Nokia chose in 1992 to focus all its energies and atten-tion on telecommunications, and over the next few yearsit exited all its other businesses, including ones withglobal scale and good performance. As deregulation andprivatization opened telecommunications services mar-kets in Europe, Nokia allied with the upstart newentrants that were challenging the established nationalmonopoly service providers. Paying careful attention tothe final users, Nokia developed new features for itsphones that gave its clients points of differentiation thatthe established service providers were slow to match.This pattern of innovation was a result of Nokia’s earlyrecognition that mobile phones were a consumer prod-uct. This recognition also led it to focus on design, aim-ing for both appealing looks and ease of use, and onbranding. At the same time, Nokia was quick to recog-nize the possibilities for developing common platformsthat would allow offering a wide variety of models whilesaving on product development, procurement, andmanufacturing costs.

Nokia’s sales more than doubled year-on-year from 1992through 1995. Growth at these rates presented immenseoperational challenges, and in 1995 logistics problems led to severe difficulties in meeting demand for its latest


275

products. Meanwhile, older products sat unsold. Thefirm’s stock market valuation was cut in half. Rallyingfrom this new crisis, Nokia introduced new informationand decision support systems and stronger operating con-trols that allowed it to triple its sales revenue between 1996and 1999 while maintaining its pace of innovation.

The organizational design that Nokia adopted was cru-cial to this success. In the early days, employees throughoutthe firm were motivated by the desire to save the companyand then to build it to be something special for themselvesand their country. The successes they collectively achievedwere a source of real, personal pride. The vision firstoffered by the leadership—that voice would go wireless—was an additional source of inspiration, as, later, was theannounced intention to supplant Motorola as the industryleader. The excitement and fun of being part of a rapidlygrowing, internationally successful firm further strength-ened employees’ identification with the company.

Other aspects of the organizational design supportedthis identification and the motivation it provided. Thearchitecture was kept fluid, with project teams forming anddissolving easily, so all had the opportunity to work oninteresting things and to build networks across the firm.Growth meant there were potentially lots of opportunitiesfor learning and taking on new responsibilities. The com-pany encouraged taking these by posting all job openingsinternally and prohibiting the bosses of those who wantedto move from blocking transfers. The leaders of the com-pany were open and very approachable, eating in the samecafeteria as the other employees. They clearly operated asa team and set an example of teamwork throughout theorganization. “Value-based leadership,” rather than control


276

through rigorous processes, was the model, and the valuesof customer satisfaction, respect for the individual, achieve-ment, and continuous learning were acted upon consistently.

Pay differences across the organization were muted.Bonuses were small and were typically paid on a teambasis and on overall company performance, not individu-ally. At the same time these low-powered explicit incen-tives were offered, stretch targets were set that shapedexpectations and helped create norms of hard work andperformance delivery. Meanwhile, innovation and experi-mentation were encouraged. More than a third of theemployees were in R&D positions, but anywhere in theorganization, if someone had an idea, it would be hard tofind anyone who would block its being tried. Honest fail-ures were not punished—for example, no one was firedover the 1995 logistics breakdown—and the leadershipsought to remove fear from the organization, so that peo-ple would be willing to take risks. Politics and influenceactivities were strongly discouraged, and a norm of open,honest communication and debate was fostered. People atNokia felt they could trust one another and their bosses,and this too made taking risk-taking easier.

In 1998 and 1999, Nokia made an even greater commit-ment to innovation, declaring its intent to lead the devel-opment of the “Mobile Information Society” that wouldput full access to the Internet on everyone’s mobile phone.Nokia’s previous innovations had been undertaken in thecontext of the externally established standards governingmobile voice communications. In contrast, no quasi-governmental body would set standards for the mobileInternet. Moreover, while mobile telephony in the 1990shad been largely concerned with voice communications,


277

data and multimedia would likely drive the next generationof mobile communications. What constituted a bettermobile phone for voice communications was prettyclear—smaller, better sound quality, longer battery life.Data communications involved much wider possibilitiesfor developing services that could be delivered over thephone, from the provision of location-sensitive informa-tion tied to the Global Positioning Satellite system tointeractive games and mobile commerce. Additionally,there were emerging technologies for wireless communi-cations apart from the phone. Thus, the company’s newinnovation efforts could, and would have to be, muchmore far ranging and open-ended than before.

To deal with this need for more exploration, Nokiain 1998 established a separate unit, Nokia VenturesOrganization (NVO). Its remit was to develop businessesthat involved combining new technologies with newmarkets. Explicitly, the two existing core business units sup-plying mobile phones (Nokia Mobile Phones—NMP) andnetwork equipment were charged with developing opportu-nities that extended existing technology into new markets orthat sought to serve existing markets with new technologies.At the same time, responsibilities were reallocated amongthe group who together had led the corporation since 1992.The president, Jorma Ollila, became chairman while retain-ing the role of CEO. Pekka Ala-Pietilä, the head of NMP,was named to replace Ollila as president of the corporationand then assigned to head the fledgling NVO and the cor-porate Central Research Laboratories. Thus, the second-highest ranking executive took responsibility for drivingexploration and growth. Meanwhile, the head of the net-works business moved over to lead NMP.


278

NVO contained both internal venture units and a cor-porate venture capital fund, which was based in SiliconValley. NVO was to function as an incubator for newbusinesses, which would, if they succeeded, return to thecore business units, become new freestanding businessunits on their own, or be sold or spun off as independentcompanies. No business would stay in NVO indefinitely.This was to limit internal resistance and jealousies thatmany companies have experienced when they set upexploratory units. Ideas for new ventures might ariseanywhere in the company. Whether they were moved intoNVO or stayed in one of the core businesses, which hadtheir own internal venturing units, was determined bythe rule of whether the ideas involved both new technologies and new markets. Teams of engineers regu-larly moved between the operating businesses, NVO, andthe Labs, so that narrowly defined interests wereavoided. Governance of NVO was by a group involvingthe leaders of the operating businesses. This helpedfurther ensure that NVO would be connected to themain businesses and not isolated from the corporation’scenter.

The collapse of the boom in telecommunications in2000 has slowed adoption of new technologies by the tele-phone service operators and has been disastrous for manyequipment suppliers. Still Nokia has continued to prosper.It has developed a range of new products and softwareplatforms, both on its own and through consortia, spend-ing over 10 percent of net sales on R&D. Meanwhile it hasincreased its market share to near 40 percent in mobilephones and maintained remarkable margins while itscompetitors are losing money.


279

Ollila attributed Nokia’s success to the company’sability to balance exploration and exploitation: “Whyhave we been a successful company? If you want a verysimple answer, it is getting the balance right betweeninnovation and execution.”4

Ollila’s answer is simple, but actually achieving thisbalance is very hard. The key is good organizational design.

Notes

1. See, for example, Saloner, Shepard, and Podolny (2001:271–84).

2. See Grinblatt and Titman (2002: 707–8) for a discussion ofthe literature in this area.

3. See Day et al. (2001) for a discussion of this architecturalsolution and Burgelman (1984) for a broad analysis of alter-native models for organizing new business developmentinitiatives.

4. Quoted in Doornik and Roberts (2002).


280

281

7

Creating the Modern Firm:Management and Leadership

Challenges

This book has offered a set of frameworks, concepts, andtools to help meet the challenges of designing effectiveorganizations. It has also offered a number of examples offirms that have developed strategies and organizationaldesigns that allowed them to deliver exceptional perfor-mance and growth. Yet, the fact remains that it is verydifficult to develop a winning strategy and an effectiveorganization. Doing so is fundamentally a creative act thatrequires both the analytical problem-solving that charac-terizes management and the vision, communication, andpersuasion that are essential features of leadership.1

For success, the elements of the firm’s strategy andorganization must be aligned with one another and mustfit the environment in which the firm operates. This needfor alignment and fit, combined with the many interde-pendencies among the elements of the strategy and orga-nizational design, means the strategy and organizationreally must be developed in tandem, in a holistic fashion.

Structure does not follow strategy any more than strategyfollows structure. This simultaneity implies that theproblem of creating a strategy and an organization thatwill allow the firm to succeed is immensely complex,because it involves so many dimensions that interact withone another.

Many of these interactions are dynamic. Especially for alarge firm, a change in strategy can easily affect the indus-try and bring a reaction that leads to the need for furtherchanges in strategy and organization. The same is trueinternally: Changes in one aspect of the organization aimedat effecting one particular change in behavior can alterother aspects of behavior in ways that necessitate furtherchanges on other dimensions of the design. Thus, the usualapproach to fixing the problems that arise as organizationsevolve—find an intervention whose first-order effect is tosolve the problem, take everything else as given, and pullthe lever—is fundamentally flawed. It only sets off a poten-tially unending stream of response, intervention, furtherunanticipated response, and yet another intervention.

Instead, strategic and organizational choices must bemade holistically, recognizing the interdependencies. Thescope of the firm—what it is going to do, where, how, andfor whom—must be decided. How it is going to distin-guish itself from the competition, gain competitive advan-tage, and create value must be determined. The rightpeople must be attracted, retained, and assigned to differ-ent roles. The formal architecture must be crafted to alloweffective coordination and motivation of these people. Theprocesses, procedures, and routines that guide and controlbehavior must be developed. The fundamental values,beliefs, and norms that will be shared across the firm must

Creating the Modern Firm

282

be created, transmitted, and adopted. And all these mustmesh properly with one another, so the organization reallydoes allow the strategy to be executed.

The people, the networks among them, and the rou-tines they follow must give the firm the capabilities itneeds to create value. The system of incentives must moti-vate the particular people that have been attracted to thefirm to supply the right mix of behaviors needed to deliveron the strategy and let the firm reach its goals. The formalstructure and the allocation of decision authority need tobe aligned with where expertise lies and with what moti-vates the people in the organization. Finally, all the ele-ments of the strategy and organization need to fit with thecompetitive, technological, social, legal, and regulatoryrealities the firm faces.

Then, as the world and the organization itself evolve,the fit must be maintained by adjusting—or even radicallychanging—the strategy and the organizational design.

Recognizing the complementarity and substitutionrelationships that exist among the design variables canhelp suggest the shape of potentially coherent patterns,and thus reduce the design problem’s complexity. Still, itremains very hard. People throughout the firm must beinvolved, because in a firm of any size the knowledge ofhow things really work, how customers really behave, howchoices really interact, is highly dispersed. Managersthroughout the firm must participate in the design task,completing the details of the parts of the strategy andorganization they know best while cooperating to ensurethat the overall result remains coherent.

Solving the organizational design problem then funda-mentally requires both management and leadership.


283

Much of the actual design work is management—puttingtogether budgeting processes, specifying reporting rela-tionships, determining what will be outsourced, settingup governance procedures, creating and staffing depart-ments, establishing the financing model. This is vitallyimportant, but it is not enough. Leadership is needed too.Leaders offer direction and then motivate others tobelieve and to follow. The basic conception of the strategyand organizational design is thus a matter of leadership.The leaders must provide a vision of the strategy andorganization, indicating the underlying principles andhow the basic trade-offs are to be resolved. They alsoneed to communicate the model in a clear and compellingway, so that others understand and embrace it and aremotivated to try to realize it in designing their parts of theorganization.

Leadership is involved in a second way in solving thedesign problem. While managers across the firm can craftand institute the formal aspects of the design, they cannotdirectly control the networks and culture. The people inthe organization, individually and collectively, determinewhat they are going to believe, what they will value, whatbehavioral norms they will adopt, and with whom theywill connect informally. Yet, these features may be themost important ones for determining behavior, and thushow well the firm performs. Leadership can shape thesechoices.

The formal elements of the design can influence thenetworks and culture, so managers can have some indirectcontrol over them. For example, BP’s establishing peergroups helped create networks that were important evenafter the original members had moved on to new jobs and


284

were no longer in the same peer group. Nokia’s choosingnot to fire anyone after its logistics disaster in 1995 helpedcreate the absence of fear in the organization that stronglypromoted people’s willingness to take risks. Still, leadershipmust play a crucial role in successfully shaping the culture.

Enunciating corporate values is the easy part and, on itsown, not very effective. Most statements of values are toovague and too abstract to have much impact. Organizationsall say that they care about their customers, but what doesthat mean in terms of behavior? Leaders must give specificmeaning to the values, which then sets the basis for theirgenerating norms of expected behavior.

The first step is for the leaders themselves to act inaccordance with the values and to model the sort of behav-ior that is wanted. A CEO who personally handles a sam-pling of customer complaints on a regular basis isindicating in a forceful way how important customer carereally is. The leaders should also celebrate and reward thosewho act in appropriate ways and correct those who do not.

Stories can be an especially powerful tool for commu-nicating what is wanted and thus for shaping what hap-pens. For example, the 3M CEO’s telling stories about thescientist who ignored his orders to drop a project andbrought it through to a successful product moulds theculture by signaling very clearly what is important. Evenmore powerful is the story told every new employee atNordstrom, the U.S. retailer renowned for its customerservice. A customer carrying a very worn set of tire chainsapproached a clerk and complained that the chains hadnot proven satisfactory. Although the customer had noreceipt, the clerk immediately refunded the claimedpurchase price without question. This was despite the


285

fact that Nordstrom has never sold tire chains, or anyother kind of automobile supplies!

Thus, organizational design involves both managementand leadership. Beyond that, it is fundamentally a creativeprocess. To succeed, a firm must create value and keepsome of it. This can happen only if the firm’s strategy andorganization together allow the firm to be better than thecompetition, to offer products or services that meet itstarget customers’ needs more effectively or more cheaply.A firm that does the same things in the same ways as thecompetition cannot be better than its rivals, and the head-to-head competition that will ensue will guarantee it getsto keep very little of any value it might create.

This means that inherently there must be somethingdistinctively different about a successful firm’s strategyand organization. For this reason, solving the problems ofstrategy and organization is an act of real creativity. Itinvolves finding something new and different that works.

Clearly, much of this creativity can take the form ofputting existing things together in novel ways. Not everyelement of the strategy and organizational design has tobe absolutely new and unique, and there is much to belearned from experience. But chasing after apparent bestpractices is largely futile if the aim is to achieve distinc-tiveness. At best it may make the firm as good as the com-petition, and this is not enough to win. The more likelyoutcome is that it results in a monstrosity, a patchwork ofill-fitting organizational features that do not add up to acoherent design. This is a recipe for failure.

Creativity involves originality, imagining new things,seeing new patterns and connections. Yet, as important asthis originality is, it is not enough. For the point is not just


286

to come up with something new, but instead somethingdistinctive that works. For this, understanding of the fun-damental logics governing organizational design isrequired. The ideas and examples offered in this book aremeant to be a start on providing this understanding.

Note

1. See Kotter (1990) on this differentiation between manage-ment and leadership.


287

289

References

Abreu, D., Pearce, D., and Stacchetti, E. (1990). “Towards aTheory of Discounted Repeated Games with ImperfectMonitoring.” Econometrica, 58: 1041–63.

Aghion, P., and Tirole, J. (1997). “Formal and Real Authorityin Organizations.” Journal of Political Economy, 105: 1–29.

Akerlof, G. A. (1970). “The Market for Lemons: QualityUncertainty and the Market Mechanism.” Quarterly Journalof Economics, 89: 488–500.

Alchian, A., and Demsetz, H. (1972). “Production, InformationCosts, and Economic Organization.” American EconomicReview, 62: 777–95.

Anderson, E. (1985). “The Salesperson as Outside Agent orEmployee: A Transactions Cost Analysis.” MarketingScience, 4: 234–54.

——, and Schmittlein, D. C. (1984). “Integration of the SalesForce: An Empirical Examination.” The RAND Journal ofEconomics, 15: 385–95.

Arrow, K. (1974). The Limits of Organization. New York: W. W. Norton & Company.

Asanuma, B. (1989). “Manufacturer–Supplier Relationships inJapan and the Concept of Relation-Specific Skill.” Journalof the Japanese and International Economies, 3: 1–30.

References

290

Asanuma, B., and Kikutani, T. (1992). “Risk Absorption inJapanese Subcontracting: A Microeconometric Study on theAutomobile Industry.” Journal of the Japanese and Inter-national Economies, 6: 1–29.

Athey, S., and Roberts, J. (2001). “Organizational Design:Decision Rights and Incentive Contracts.” American EconomicReview: Papers and Proceedings, 91: 200–5.

Avery, C., Chevalier, J. A., and Schaefer, S. (1998). “Why DoManagers Undertake Acquisitions?: an Analysis of Internaland External Rewards to Acquisitiveness.” Journal of Law,Economics, and Organization, 14: 24–43.

——, Roberts, J., and Zemsky, P. (1993). “Sony CorporationEnters the Entertainment Business.” Stanford, CA: StanfordUniversity Graduate School of Business, case S-BP-265.

Baker, G. (2000). “The Use of Performance Measures inIncentive Contracting.” American Economic Review: Papersand Proceedings, 90: 415–20.

——, Gibbons, R., and Murphy, K. (1994). “SubjectivePerformance Measures in Optimal Incentive Contracts.”Quarterly Journal of Economics, 109: 1125–56.

—— —— —— (2001). “Relational Contracts and the Theoryof the Firm.” Quarterly Journal of Economics, 117: 39–83.

Baron, J., and Kreps, D. M. (1999). Strategic Human Resources:Frameworks for General Managers. New York: John Wiley &Sons.

——, Burton, D., and Hannan, M. T. (1996). “The RoadTaken: Origins and Early Evolution of EmploymentSystems in Emerging Companies.” Industrial and CorporateChange, 5: 239–76.

Barnett, W., and Reddy, P. (1995). “Newell Company (A).”Stanford, CA: Stanford University Graduate School ofBusiness, case S-SM-16A.

Bartlett, C. A. (1993). “ABB’s Relays Business: Buildingand Managing a Global Matrix.” Boston: Harvard

References

291

University Graduate School of Business Administration,case 9-394-016.

——, and Mohammed, A. (1995). “3M: Profile of anInnovating Company.” Boston: Harvard UniversityGraduate School of Business Administration, case 9-395-016.

——, and O’Connell, J. (1998). “Lincoln Electric: VenturingAbroad.” Boston: Harvard University Graduate School ofBusiness Administration, case 3-398-095.

Berg, N. A., and Fast, N. D. (1975). “Lincoln Electric Co..”Boston: Harvard University Graduate School of BusinessAdministration, case 9-376-028.

Berger, P., and Ofek, E. (1995). “Diversification’s Impact onFirm Value.” Journal of Financial Economics, 37: 39–65.

—— —— (1996). “Bustup Takeovers of Value-DestroyingDiversified Firms.” Journal of Finance, 51: 1175–200.

Berzins, A., Podolny, J., and Roberts, J. (1998a). “BritishPetroleum (A): Performance and Growth.” Stanford, CA:Stanford University Graduate School of Business, case S-IB-16A.

—— —— —— (1998b). “British Petroleum (B): Focus onLearning.” Stanford, CA: Stanford University GraduateSchool of Business, case S-IB-16B.

Brady, D., and de Verdier, A.-K. (1998). “Nike: A History.”Stanford, CA: Stanford University Graduate School ofBusiness, case S-IB-14A.

Bresnahan, T., Brynjolfsson, E., and Hitt, L. M. (2002).“Information Technology, Workplace Organization and theDemand for Skilled Labor: Firm-level Evidence.” QuarterlyJournal of Economics, 117: 339–76.

Brynjolfsson, E., and Hitt, L. M. (2000). “Beyond Computation:Information Technology, Organizational Transformation andBusiness Performance.” Journal of Economic Perspectives,14/4: 23–48.

References

292

Burgelman, R. (1984). “Designs for Corporate Entrepreneurshipin Established Firms.” California Management Review, 26:154–66.

—— (2002). Strategy is Destiny: How Strategy-Making Shapesa Company’s Future. New York: Free Press.

Burt, T. (2002). “Auditors Drive a Hard Bargain at Ford.”Financial Times, January 14, U.S. edition, 14.

Campa, H., and Kedia, S. (2002). “Explaining theDiversification Discount.” Journal of Finance, 57: 1731–62.

Chandler, A., Jr. (1962). Strategy and Structure. Cambridge,MA: MIT Press.

—— (1977). The Visible Hand: The Managerial Revolution inAmerican Business. Cambridge, MA: Belknap Press.

Chevalier, J. (2002). “Why do Firms Undertake DiversifyingMergers? An Examination of the Investment Policies ofMerging Firms.” Chicago: University of Chicago GraduateSchool of Business, working paper.

Coase, R. (1937). “The Nature of the Firm.” Economica, 4:386–405.

—— (1960). “The Problem of Social Cost.” Journal of Lawand Economics, 3: 1–44.

Comment, R., and Jarrell, G. (1995). “Corporate Focus and Stock Returns.” Journal of Financial Economics, 37:67–87.

Day, J., Mang, P., Richter, A., and Roberts, J. (2001). “TheInnovative Organization: Why New Ventures Need Morethan a Room of their Own.” The McKinsey Quarterly,(Second Quarter): 20–31.

—— —— —— —— (2002). “Has Performance Pay Hadits Day?” The McKinsey Quarterly (Fourth Quarter): 46–55.

Doornik, K. (2001). “Relational Contracting in Partnerships.”Stanford, CA: Stanford University Graduate School ofBusiness, working paper.

References

293

—— (2002). “Incentive Contracts with Dispute Costs.”Stanford, CA: Stanford University Graduate School ofBusiness, working paper.

—— (2003). “Dispute Costs and Reputation.” Oxford:Oxford University Saïd School of Business, working paper.

Doornik, K., and Roberts, J. (2001). “Nokia Corporation:Innovation and Efficiency in a High-Growth Global Firm.”Stanford, CA: Stanford University Graduate School ofBusiness, case S-IB-23.

Gibbons, R., (1997). “Incentives and Careers inOrganizations.” In D. Kreps and K. Wallis (eds.), Advancesin Economic Theory and Econometrics, vol. II. Cambridge:Cambridge University Press, 1–37.

——, and Murphy, K. (1990). “Relative PerformanceEvaluation for Chief Executive Officers.” Industrial andLabor Relations Review, 43: 30S–51S.

—— —— (1992). “Optimal Incentive Contracts in thePresence of Career Concerns: Theory and Evidence.”Journal of Political Economy, 100: 468–505.

Gibbons, R., and Waldman, M. (1999). “Careers inOrganizations: Theory and Evidence.” In O. Ashenfelterand D. Card (eds.), Handbook of Labor Economics, vol. 3B.Amsterdam: Elsevier, 2373–437.

Gomes-Casseres, B., and McQuade, K. (1991). “Xerox andFuji Xerox.” Boston: Harvard University Graduate Schoolof Business Administration, case 9-391-156.

Grinblatt, M., and Titman, S. (2002). Financial Markets andCorporate Strategy. Boston: McGraw-Hill Irwin.

Grossman, S. J., and Hart, O. (1986). “Costs and Benefits ofOwnership: A Theory of Vertical and Lateral Integration.”Journal of Political Economy, 94: 691–719.

Hart, O. (1995). Firms, Contracts, and Financial Structure.Oxford: Clarendon Press.

References

294

Hart, O., and Holmström, B. (1987). “The Theory ofContracts.” In T. Bewley (ed.), Advances in EconomicTheory: Fifth World Congress. Cambridge: CambridgeUniversity Press, 71–155.

——, and Moore, J. (1990). “Property Rights and the Natureof the Firm.” Journal of Political Economy, 98: 1119–58.

Hellmann, T. (1998). “The Allocation of Control Rights inVenture Capital Contracts.” RAND Journal of Economics,29: 57–76.

Helper, S., MacDuffie, J. P., and Sabel, C. (1998). “TheBoundaries of the Firm as a Design Problem.” Proceedings,Conference on Make versus Buy: Emerging Structures.New York: Columbia University School of Law SloanProject in Corporate Governance.

Holmström, B. (1979). “Moral Hazard and Observability.”Bell Journal of Economics, 10: 74–91.

—— (1982a). “Managerial Incentive Problems—A DynamicPerspective.” In Essays in Economics and Management inHonor of Lars Wahlbeck. Helsinki: Swedish School ofEconomics. Reprinted in Review of Economic Studies, 66(1999): 169–82.

—— (1982b). “Moral Hazard in Teams.” Bell Journal ofEconomics, 13: 324–40.

—— (1999). “The Firm as a Subeconomy.” Journal of Law,Economics, and Organization, 15: 74–102.

——, and Milgrom, P. (1991). “Multitask Principal-AgentAnalyses: Incentive Contracts, Asset Ownership and JobDesign.” Journal of Law, Economics, and Organization, 7:24–52.

——, and Roberts, J. (1998). “The Boundaries of the FirmRevisited.” Journal of Economic Perspectives, 12: 73–94.

Horngren, C. T. (1999). Cost Accounting: A ManagerialEmphasis, 5th edn. Englewood Cliffs, NJ: Prentice Hall.

References

295

Ichniowski, C., Shaw, K., and Prennushi, G. (1997). “TheEffects of Human Resource Management Practices onProductivity: A Study of Steel Finishing Lines.” AmericanEconomic Review, 87: 291–313.

Jaikumar, R. (1986). “Postindustrial Manufacturing.” HarvardBusiness Review, 64: 61–8.

Jensen, M., and Meckling, W. (1976). “Theory of the Firm:Managerial Behavior, Agency Costs and OwnershipStructure.” Journal of Financial Economics, 3: 305–60.

Joskow, P. L. (1985). “Vertical Integration and Long TermContracts: The Case of Coal Burning Electric GeneratingPlants.” Journal of Law, Economics, and Organization, 1: 33–80.

—— (1987). “Contract Duration and Relationship SpecificInvestments: Empirical Evidence from Coal Markets.”American Economic Review, 77: 168–85.

—— (1988). “Asset Specificity and the Structure of VerticalRelationships: Empirical Evidence.” Journal of Law,Economics, and Organization, 4: 95–117.

Kamper, A., Podolny, J., and Roberts, J. (2000). “NovoNordisk: Global Coordination.” Stanford, CA: StanfordUniversity Graduate School of Business, case S-IB-20A.

Kawasaki, T., and McMillan, J. (1987). “The Design ofContracts: Evidence from Japanese Subcontracting.” Journalof the Japanese and International Economies, 1: 327–49.

Kennan, J., and Wilson, R. (1993). “Bargaining with Private Information.” Journal of Economic Literature, 31:45–104.

Klein, B., Crawford, R., and Alchian, A. A. (1978). “VerticalIntegration, Appropriable Rents, and the CompetitiveContracting Process.” Journal of Law and Economics, 21:297–326.

Kotter, J. (1990). “What Leaders Really Do.” Harvard BusinessReview, 68: 103–11.

References

296

Kreps, D. (1990). “Corporate Culture and EconomicTheory.” In J. Alt and K. Shepsle (eds.), Perspectives onPositive Political Economy. Cambridge: CambridgeUniversity Press, 90–143.

Lang, L. H. P., and Stulz, R. M. (1994). “Tobin’s q, CorporateDiversification and Firm Performance.” Journal of PoliticalEconomy, 102: 1248–80.

Lazear, E. (2000). “Performance Pay and Productivity.”American Economic Review, 90: 1346–61.

——, and Rosen, S. (1981). “Rank-Order Tournaments asOptimum Labor Contracts.” Journal of Political Economy,89: 841–64.

Levin, J. (2003). “Relational Incentive Contracts.” AmericanEconomic Review, 93: 835–57.

Levinthal, D. A. (1997). “Adaptation on Rugged Landscapes.”Management Science, 43: 934–50.

Lichtenberg, F. (1992). “Industrial De-Diversification and itsConsequences for Productivity.” Journal of EconomicBehavior and Organization, 18: 427–38.

Lins, K., and Servaes, H. (1999). “International Evidence onthe Value of Corporate Diversification.” Journal of Finance,54: 2215–40.

McMillan, J. (1995). “Reorganizing Vertical SupplyRelationships.” In H. Siebert (ed.), Trends in BusinessOrganization: Do Participation and Cooperation IncreaseCompetitiveness? Tübingen: J. C. B. Mohr, 203–22.

—— (2002). Reinventing the Bazaar: The Natural History ofMarkets. New York: W. W. Norton & Company.

——, and Woodruff, C. (1999a). “Dispute Prevention WithoutCourts in Vietnam.” Journal of Law, Economics, andOrganization, 15: 637–58.

—— —— (1999b). “Interfirm Relationships and InformalCredit in Vietnam.” Quarterly Journal of Economics, 114:1285–20.

References

297

March, J. (1991). “Exploration and Exploitation inOrganizational Learning.” Organization Science, 2: 71–87.

Masten, S. E. (1984). “The Organization of Production:Evidence from the Aerospace Industry.” Journal of Law andEconomics, 27: 403–17.

——, Meehan, J. W., and Snyder, E. A. (1989). “VerticalIntegration in the US Auto Industry: A Note on theInfluence of Transactions Specific Assets.” Journal ofEconomic Behavior and Organization, 12: 265–73.

Matsusaka, J. (1993). “Takeover Motives During theConglomerate Merger Wave.” RAND Journal of Economics,24: 357–79.

Meyer, M., Milgrom, P., and Roberts, J. (1992). “OrganizationalProspects, Influence Costs and Ownership Changes.” Journalof Economics and Management Strategy, 1: 9–36.

Milgrom, P., and Roberts, J. (1988a). “Communication andInventories as Substitutes in Organizing Production.”Scandinavian Journal of Economics, 90: 275–89.

—— —— (1988b). “An Economic Approach to InfluenceActivities in Organizations.” American Journal of Sociology,94 Suppl.: S154–79.

—— —— (1990a). “Bargaining Costs, Influence Costs andthe Organization of Economic Activity.” In J. Alt and K. Shepsle (eds.), Perspectives on Positive Political Economy.Cambridge: Cambridge University Press, 57–89.

—— —— (1990b). “The Economics of ModernManufacturing: Technology, Strategy and Organization.”American Economic Review, 80: 511–28.

—— —— (1990c). “The Efficiency of Equity in OrganizationalDecision Processes.” American Economic Review: Papers andProceedings, 80: 154–9.

—— —— (1992). Economics, Organization and Management.Englewood Cliffs, NJ: Prentice Hall, 1992.

References

298

Milgrom, P., and Roberts, J. (1993). “Johnson Controls, Inc.,Automotive Systems Group: Georgetown, KentuckyPlant.” Stanford, CA: Stanford University Graduate Schoolof Business, case S-BE-9.

—— —— (1994). “Complementarities and Systems:Understanding Japanese Economic Organization.” EstudiosEconomicos, 9: 3–42.

—— —— (1995). “Complementarities and Fit: Strategy,Structure and Organizational Change in Manufacturing.”Journal of Accounting and Economics, 19: 179–208.

—— —— (1998). “The Internal Politics of the Firm.” In S. Bowles, M. Franzini, and U. Pagano (eds.), The Politics ofExchange and the Economics of Power. London: Routledge,46–62.

Monteverde, K., and Teece, D. (1982). “Supplier SwitchingCost and Vertical Integration in the U.S. AutomobileIndustry.” Bell Journal of Economics, 13: 206–13.

Montgomery, C. A., and Wernerfelt, B. (1988). “Diversification,Ricardian Rents and Tobin’s q.” RAND Journal of Economics,19: 623–32.

Myerson. R., and Satterthwaite, M. (1983). “EfficientMechanisms for Bilateral Trading.” Journal of EconomicTheory, 29: 265–81.

Nagar, V. (2002). “Delegation and Incentive Compensation.”Accounting Review, 77: 379–95.

Newman, P. C. (1985). Company of Adventurers, Vol. I.Markham, ON: Penguin.

—— (1987). Company of Adventurers, Vol. II: Caesars of theWilderness. Markham, ON: Penguin.

—— (1991). Company of Adventurers, Vol. III: MerchantPrinces. Markham, ON: Penguin.

Nickerson, J. and Zenger, T. (2003). “Being Efficiently Fickle:A Dynamic Theory of Organizational Choice.” OrganizationScience, 13: 547–66.

References

299

Ohno, T. (1988). Toyota Production System: Beyond Large-Scale Production. Cambridge, MA: Productivity Press.

O’Reilly III, C. (1998). “Cisco Systems: The Acquisition ofTechnology is the Acquisition of People.” Stanford, CA:Stanford University Graduate School of Business, case S-HR-10.

Oyer, P. (1998). “Fiscal Year Ends and Nonlinear IncentiveContracts: The Effect on Business Seasonality.” QuarterlyJournal of Economics, 113: 149–85.

Pearson, A., and Hurstak, J. (1992). “Johnson and Johnson: Hospital Services.” Boston: Harvard UniversityGraduate School of Business Administration, case 9-392-050.

Pfeffer, J. (1996). Competitive Advantage Through People:Unleashing the Power of the Work Force. Boston: HarvardBusiness School Press.

Porter, M. (1980). Competitive Strategy: Techniques ForAnalyzing Industries And Competitors. New York: Free Press.

—— (1985). Competitive Advantage: Creating And SustainingSuperior Performance. New York: Free Press.

Prendergast, C. (1999). “The Provision of Incentives inFirms.” Journal of Economic Literature, 37: 7–63.

—— (2000). “What Trade-off of Risk and Incentives?” AmericanEconomic Review: Papers and Proceedings, 90: 421–5.

Rajan, R., Servaes, H., and Zingales, L. (2000). “The Cost ofDiversity: The Diversification Discount and InefficientInvestment.” Journal of Finance, 60: 35–80.

——, and Wulf, J. (2002). “The Flattening Firm: Evidence from Panel Data on the Changing Nature of Corporate Hierarchies.” Philadelphia: University of Pennsylvania Wharton School of Finance, working paper.

——, and Zingales, L. G. (1998). “Power in a Theory of theFirm.” Quarterly Journal of Economics, 113: 387–432.

References

300

Roberts, J. (1998). “Value Maximization.” In P. Newman (ed.),The New Palgrave Dictionary of Economics and the Law,vol. 3. London: Macmillan Reference.

——, and Van den Steen, E. (2001). “Human Capital andCorporate Governance.” In J. Schwalbach (ed.), CorporateGovernance: A Volume in Honor of Horst Albach. Berlin:Springer Verlag, 128–44.

Rotemberg, J., and Saloner, G. (1994). “The Benefits ofNarrow Business Strategies.” American Economic Review,84: 1330–49.

—— —— (2000). “Visionaries, Managers and StrategicDirection.” RAND Journal of Economics, 31: 693–716.

Roy, D. (1952). “Quota Restriction and Goldbricking ina Machine Shop.” American Journal of Sociology, 57:427–42.

Ruigrok, W., Pettigrew, A., Peck, S., and Whittington, R. (1999).“Corporate Restructuring and New Forms of Organizing:Evidence from Europe.” Management International Review,39 (Special Issue): 41–64.

Saloner, G., Shepard, A., and Podolny, J. (2001). StrategicManagement. New York: John Wiley & Sons.

Schaefer, S. (1988). “Influence Costs, Structural Inertia andOrganizational Change.” Journal of Economics andManagement Strategy, 7: 237–63.

Scharfstein, D. (1998). “The Dark Side of Internal CapitalMarkets II: Evidence from Diversified Conglomerates.”Cambridge, MA: National Bureau of Economic Research,working paper 6352.

——, and Stein, J. (2000). “The Dark Side of Internal CapitalMarkets: Divisional Rent-Seeking and InefficientInvestment.” Journal of Finance, 55: 2537–64.

Shin, H.-H., and Stultz, R. (1998). “Are Internal Capital MarketsEfficient?” Quarterly Journal of Economics, 112: 531–52.

Simon, H. (1951). “A Formal Theory of the EmploymentRelationship.” Econometrica, 19: 293–305.

References

301

—— (1991). “Organizations and Markets.” Journal ofEconomic Perspectives, 5: 25–44.

Smith, A. (1776/1937). In E. Canaan (ed.), An Inquiry into theNature and Causes of the Wealth of Nations. New York: TheModern Library.

Spence, A. M. (1973). “Job Market Signaling.” Quarterly Journalof Economics, 87: 355–74.

Spraakman, G. P. (2002). “A Critique of Milgrom and Roberts’Treatment of Incentives vs. Bureaucratic Controls in theBritish North American Fur Trade.” Journal of ManagementAccounting Research, 14: 135–52.

Stevenson, H., Martinez, J., and Jarillo, J. C. (1989). “BenettonS.p.A.” Boston: Harvard University Graduate School ofBusiness Administration, case 9-389-074.

Vance, R., Bhambri, A., and Wilson, J. (1980). “IBM Corp.: TheBubble-Memory Incident.” Boston: Harvard UniversityGraduate School of Business Administration, case 9-180-042.

Van den Steen, E. (2002). “Organizational Beliefs andManagerial Vision.” Cambridge, MA: Massachusetts Instituteof Technology Sloan School of Management, working paper.

Villalonga, B. (2002a). “Diversification Cost or Premium?New Evidence from BITS Establishment-Level Data.”Boston: Harvard University Graduate School of BusinessAdministration, working paper.

—— (2002b). “Does Diversification Cause the ‘DiversificationDiscount’?” Boston: Harvard University Graduate Schoolof Business Administration, working paper.

Whang, S., and de Verdier, A.-K. (1998). “Nike: Global SupplyChain.” Stanford, CA: Stanford University GraduateSchool of Business, case S-IB-14D.

Whinston, M. (2003). “On the Transaction Cost Determinantsof Vertical Integration.” Journal of Law, Economics, andOrganization, 19: 1–23.

Whittington, R., Pettigrew, A., Peck, S., Fenton, E., andConyon, M. (1999). “Change and Complementarities in the

New Competitive Landscape: A European Panel Study,1992–1996.” Organization Science, 10: 583–600.

Williamson, O. (1975). Markets and Hierarchies: Analysis andAntitrust Implications. New York: Free Press.

—— (1985). The Economic Institutions of Capitalism. NewYork: Free Press.

Womack, J. P., Jones, D. T., and Roos, D. (1990). The Machinethat Changed the World. New York: Harper Perennial.

References

302

Index

Figures, tables, and notes are indicated by f, t, and n respectively.

3M 258–60, 267–8, 285incentives in 267–8innovations in 268niche markets in 260performance measures 259rewards in 259subjective evaluations in 259

ABB Asea Brown Boveri 101, 109,239–40

ABACUS reporting system 167and small business units 166matrix forms 239organizational form 109–10strategy in 110–11

Abreu, D. 200accounting returns 137–8acquisitions 166, 214, 222, 224, 227–9,

245, 245–53, 249, 151and focus 228foreign in Lincoln Electric Company

166growth by 248–9in Nokia Corporation 224in Sony 220mergers and 214, 226, 247, 250–1

integration 251target 248–9effect on value of 248

actions, observability of 127–8adaptation to change 71adverse selection 83–4, 89agency:

economic theory of 123problems 126relationships, multi-tasking in 140

model 140–3reputations in 242 ntheory 126–36, 158–60, 164, 177 n

agents (managers):actions 127–8behavior 171performance of 149

Aghion, P. 99, 170, 233Akerlof, G. A. 82Alchian, A. A. 86, 90, 103alliances 114, 207American Telephone and Telegraph

Company 254Amoco see BPAnderson, E. 106, 149antitrust 244Apple computer 164, 273architecture of firms 166–8, 180,

240, 282elements of 180features of 17

303

architecture of firms (cont.):internal 240and motivation 168

Arco see BPArrow, K. 78, 104Asanuma, B. 205Asea Brown Boveri see ABB Asea Brown

Boveriassembly line 42, 58, 68“asset federation” in BP 184assets 91–3, 95–6, 107, 225

in BP 186–90ownership of 104pricing theory 215specific 193specificity 117 n

asymmetry of information seeinformation asymmetry

AT&T 213, 254Athey, S. 141automobile industry 38–9

in Japan 64–5Avery, C. 177 n, 229

Baker, G. 158“Balanced Scorecard” 140bargaining 85, 95

Coasian 218Barnett, W. 250Barnevik, Percy 110:

see also ABB Asea Brown BoveriBaron, J. 46, 174Bartlett, C. A. 166“basket weave” model in General

Motors (GM) 239Bell Laboratories 254Benetton 191–2Berg, N. A. 41Berger, P. 225, 226Berzins, A. 25best practices 72, 112, 286

in BP Exploration 112, 173; sharingin BP Exploration 173–4, 187–8

Betamax 219Bhambri, A. 153bonus 150, 157boundaries of firms 240

horizontal 230–1vertical 230–1

BP 30, 95, 136–7, 172–3, 182–90, 239,250–2:

see also Browne, Johnand Amoco 182, 190, 250and Arco 182, 190, 250and Burmah Castrol 182“asset federation” 184assets of 186–90

lifestage of 197best practice in 173, 187capital budgets in 184changes in culture in 189discrete business units 189employees in 184ExCo 186, 188–9federal groups in 188human resources in 188management in 183–4matrix system in 183operations in 185organizational changes 182organizational design 185–6, 190 organizational model 188 outsourcing 188pay in 238peer assists 113, 187–9peer challenges 189peer groups 113, 172–3, 187–9,

239, 284–5performance 137, 190

performance contracts 189performance reviews 186performance targets 186, 188

production in 185Quarterly Performance Reviews

186–7rationalization in 184redesign of 25

Index

304

Regional Operating Companies185–6

“self help” in 95, 136strategic changes in 182–3

BP Exploration (BPX) 24–5, 30, 112,114, 182–7, 256

best practice sharing 173–4, 187–8incentives in 172–3linkages in 239small business units 172

Brady, D. 191Bresnahan, T. 46British Petroleum see BPBritish Telecom see BTBrowne, John 25–6, 112, 187, 189Brynjolfsson, E. 37, 38, 46, 73 nBT 213Burgelman, R. 257, 273, 280 nBurmah Castrol 182Burt, T. 208Burton, D. 46Businesses units:

discrete 189small 166

buying growth see growth, buying

Campa, H. 227capital allocation, internal 223–4, 226capital goods, depreciation of 169career management 201carve-outs 114CBS 246CBS Records 218, 251

and Sony 218centralization 94centralized decision making 55CEOs’ compensation 138–9: see also

compensationChandler, A. 13, 30 n, 93changes 241, 242 n

adaptation to 71and performance 66environmental, in firms 23, 27, 50,

54, 60, 68, 71–2, 111

organizational 22–7, 51, 56, 60, 66–8,231, 237, 241, 282

environmental 111; in BP 182–3strategic 23

cheating 162, 172, 200Chevalier, J. A. 177 n, 226, 227, 229choices:

organizational 52, 54strategic 32, 52, 54

Cisco Systems 246, 250, 252CNC see computer numerically

controlled (CNC)Coase, R. 77, 81, 87, 88, 103Coasian bargaining 218: see also

bargainingColumbia Pictures 218

and CBS Records 218and Sony 218, 251

Comment, R. 214, 226commitment 233

problems of 85, 90communication 65, 259

among employees 101, 232among managers 237in the firm 217

Compaq computer 252comparative performance evaluation 152compensation 138–9, 266, 271: see also

executive compensationCEOs’ 138–9

competition 3, 114, 123imperfect 87internal 269

competitive advantage 5, 14–15competitive bidding 248complementarities 33–51, 55, 58, 60,

67–9, 143, 218, 234, 240–1,249–51, 283

and discrete coherent patterns 40 fand internal organizations 240in human resource management

(HRM) 46in Lincoln Electric Company 41in Newell Co. 250

Index

305

complementarities (cont.):of design 190, 241relationships 232, 283

computer numerically controlled (CNC)machines 41, 195

concavity 51–2, 54: see alsonon-concavity

of the objective 55Concert (AT&T and BT joint venture)

213“concurrence system” in IBM 153conflict of interest 127conglomerates 214, 217, 222–3

and stock markets 214conglomerates discounts 214, 226contracting 94–5, 218contractors 169contracts 86–7, 90–2, 94, 96, 121–3,

125, 198employment 104enforcement of 85, 122–3, 158incentive 153incomplete 115 nlong-term 117 nperformance 136relational 114, 197–8, 212renegotiation of 86–7, 92short-term 206termination of 202–3

convexity 51–2: see also non-convexitycooperation 197–8, 204, 206, 207, 213,

218and initiative 106–15, 108 f, 196and teamwork 154between firms 209measuring 114

coordination 88, 103in firms 78, 107of markets 77problems of 75–6

corporate:form 180management 120performance 155

incentives and 270

policies, clarity of 231, 237strategy 16values 285

coupling, loose see loose couplingcoupling, tight see tight couplingCrawford, R. 86, 90“creosote plant” 257, 271–2cross-subsidization 223, 225, 227culture 5, 18, 26, 28, 30, 56, 72, 172,

173, 175, 190, 220, 221, 261, 272,284–5

aspects of 34, 176elements of 56, 154in firms 18, 172norms of 180PARC 17in trust 176

current:performance 98, 168–9, 243, 246,

255, 264, 269incentives for 264

Dai-ichi Kango Bank 253Daimler-Benz 215Day, J. 273, 280 nde Verdier, A.-K. 191decentralized structure 238

in Johnson and Johnson 238decision-making 2, 28, 80, 100, 125,

150–2, 171, 233, 235, 237–8, 270centralized 55, 108decentralized 55in units 152–3limit to 102monitoring of 171

decisions 17, 28, 101–2, 152, 233,236–8, 272

authority for 17–18, 56, 141, 170,237, 283

processes for 101, 153delayering 114, 232–3, 236, 242 ndelegation 242 nDelphi see General Motors (GM), Delphidemergers 214Demsetz, H. 103

Index

306

Denso 202design: see also organizational design

“coherence” in 57–9the decision-making process 152elements of 2, 190, 191, 284of employee incentives 35, 42–43, 64,

67, 94–98, 103, 123, 125, 127–53,157–60, 165, 166, 169–72, 175, 178, 192, 194, 232, 265, 273, 283

features of 234, 241of firm 1of incentives 132of jobs 141, 172, 268of performance contracts 136of system 151techniques of 212

disaggregation 180–1in BP Exploration (BPX) 112, 181, 184in firms 181model 240

discrete business units 231–2in BP 189leaders of 232

diversification 215–17, 219, 226–30and value 226

“diversification discount” 224–6division of labor (A. Smith) 74divisions 1

performance measurement of 150–1Doornik, K. 115 n, 242 ndual reporting 110

earnings see payEastman Kodak 212economic:

activities 74–5, 77–8, 88in the firm 93

efficiency 88–90, 94, 116 n, 181, 225–6“efficiency wage” 178 n; see also pay“effort provision” 127empire building 229, 249employees 3, 169

incentives 94–5, 97–9, 107–8, 123–4,128, 131–4, 141–6, 148, 158, 165,

171, 192, 218, 224, 232–5, 255,262–4

and ownership stakes 169and tools 169–70for teachers in USA 146–7manipulation of 159quantitative measures 255

monitoring of 124, 172, 176recruitment of 164–5relationship with employer 103turnover 166

employment:contracts 104relations 48

empowerment:of managers 170of small business units 235–6,

239enforcement of contracts 122, 158Enron 156–8Ericsson 274–5evaluations, subjective see subjective

evaluationsexcludable public goods 80ExCo see BP, Excoexecutive:

compensation 137–8, 249, 271and stock price 249, 271

overload 237team 240

externalities 80, 119, 217–18

family-owned firms 215–16Fast, N. D. 41federal groups in BP 188firms 74–117; see also architecture of

firms; family-owned firms;modern firms

and markets 88–103behavior in 103, 139boundaries in 168, 240

horizontal 230–1; vertical 230–1communication in the 217coordination in 78, 107design of 1, 73

Index

307

firms (cont.):diversified 216–17, 219, 224–5, 227–9

and acquisitions 227–8discount 224literature on 225–7

environmental changes in 12–13, 18,23–4, 27, 50, 60, 68, 71–2, 111,230

external boundaries of 168family-owned 215–16financial structure of 171innovations in 253–68investments in 97–9, 117 nmultibusiness 223–4nature of 74–115norms in 56organizational design of 2, 72,

180–241ownership of 171processes in 240purpose of 74–115strategy in 72

flexibility 40, 45, 68Flextronic 192focus in firms 228, 230, 242 nFord, H. 38–9Ford Motor Company 38–9, 208, 242 nfree riding 124, 154, 160, 194,

208, 234“front-and-back” model 167Fuji Photo Film 209–12Fuji Xerox 209–12functional organization 167

Geneen, H. 217, 242 nGeneral Electric (GE) 41, 217, 227, 245

GE Capital 245General Motors (GM) 40–1, 62–3, 201,

207–8, 227, 242 n, 245, 272and suppliers 207–8and Toyota 201basket weave model 239Delphi 205organizational choices in 40–1strategy and organization 63

Gibbons, R. 160, 177 n, 242 nglobalization 230Gomes-Casseres, B. 209Grinblatt, M. 280 nGross, B. 258Grossman, S. J. 95, 96, 115 ngroup norm 154group performance pay 154Grove, A. 273growth 230, 244, 264–5, 269, 271; see

also organizing for growthbuying 245–53by acquisitions 248–9in organizations 243–80in Nokia 276strategy for 230

H. J. Heinz & Co 156Hannan, M. T. 46Hart, O. 95, 96, 115 n, 117 nHBC see Hudson Bay Company (HBC)Hellmann, T. 261Helper, S. 202Hewlett-Packard (HP) 252–3high-commitment human resource

management 135, 156, 174–5,273–4

Hitt, L. M. 37, 38, 46, 73 nhold-up 92–3, 107, 116 n, 193, 195

investments in 90Holmström, B. 104, 106, 115 n, 136,

161, 177 n, 234horizontal linkages see linkages,

horizontalhorizontal scope see scope, horizontalHorngren, C. T. 156Horton, Robert 183Houston Oil and Minerals 222–3, 252Hudson Bay Company (HBC) 4–11, 13,

15–16, 30, 60, 261–2human resource management (HRM)

46, 135–40, 174career management 201complementarity, literature on 46

Hurstak, J. 111, 181

Index

308

IBM 212, 272“concurrence system” in 153

Ichniowski, C. 46ICI 215IdeaLab 258incentives 42–3, 94–9, 101,

105–8, 114, 117 n, 123–5, 128–36, 141–51, 150–1,153, 157, 159, 165, 169–73,179 n, 197, 202, 218, 224, 232,234–5, 237, 262–4, 267, 270, 283

and corporate performance 270and piece rates 42–4contracts 153design of 132, 157employee see employees, incentivesfor cooperation 114for current performance 98, 264for exploration 266for innovation 265in BP Exploration 173in firms 105, 117 nin Houston Oil and Minerals 222in Lincoln Electric Company 144in Nokia Corporation 277in sales force 105–6, 138, 194in two-supplier policy 202misaligned 238monetary 119pay 135, 242 n

performance-based 35, 99, 170performance see performance,

incentivesschemes 130, 134, 140, 152

in Sears Roebuck 157incomplete contracts 115 n: see also

contractsindividual performance pay 154: see also

performance, payindivisibilities 56influence activities 100–2, 160, 170,

225–6influence costs 67, 102, 253, 271–2

influence techniques of 100information 193, 235, 237, 239, 247–8,

272–3about performance 158asymmetries of 82–4, 87, 89, 160,

204–5, 216–17, 272–3gathering 106, 170, 193markets for 80sharing (in Toyota) 204–5systems 232, 237technology 38, 112, 237

Informativeness Principle 136–7, 151

initiative 184and cooperation see cooperation and

initiativeinnovations 243–5, 257, 264, 269

in 3M 268in firms 253–68in Lincoln Electric Company 260in Nokia Corporation 274–5, 277in organizations 3, 63in Proctor and Gamble 266incentives for 265

Intel 256, 273intellectual property 193internal competition 269international expansion 230interventions 99–102, 237

and managers 99–100inventories 64, 68

in Toyota 65investments:

in businesses 226in firms 97–9, 116 n, 117 n, 169

ITT 217

Jaikumar, R. 41Japan:

automobile industry 64–5, 196–7,242 n

economic performance 69–70firms in 197governance system 69production lines 63

Index

309

Jarillo, J. 191Jarrell, G. 214, 226Jensen, M. 103Jobs, Steve 164jobs:

functional definition 167design of 141, 172, 268–80

for multi-tasking 268–80performances in 114design of 140–1redesign of 3, 172

Johnson and Johnson 111, 181, 238decentralized structure 238

joint ventures 115, 211Jones, D. T. 197Joskow, P. L. 116 njust-in-time 65JVC (Victor Company of Japan) 219

kaizen 54, 57Kamper, A. 113Kawasaki, T. 206Kedia, S. 227Kennan, J. 90Kikutani, T. 206Klein, B. 86, 90Kotter, J. 287 nKPMG 208Kreps, D. M. 46, 164, 174kyohokai 205

Lang, L. H. P. 214, 225Lazear, E. 165leadership 62–3, 65, 283–6

in modern firms 281–7in Sony 220–1“value-based leadership” in Nokia

Corporation 276–7vision in 63

lean manufacturing see manufacturing,lean

“lean” production systems 45learning 71Levin, J. 242 nLevinthal, D. A. 32

liability, limited 131Lichtenberg, F. 142life cycle 244Lincoln Electric Company 41–4, 144,

260–2and foreign acquisitions 166complementarities in 41competitors of 260employees

incentives in 144turnover of 166

exploitation in 255–6incentives in 144organizational design in 42, 44–5, 64,

68, 255ownership of 44pay in 166piece rates in 42–4, 64, 256production in 43quality 144specialization 260work-in-process inventory 68, 256

Lincoln, James F. 64:linkages 239

among staff, horizontal 232between performances 240horizontal 232, 241in BPX 239

Lins, K. 225local optimality see optimality, localLockheed 271–2lock-in 90–1long-term contracting 117 n: see also

contractsloose coupling 67–3, 181, 237: see also

tight coupling

MacDuffie, J. P. 202make-buy choice 195make-to-order 35make-to-stock 35management 108, 283–4, 286

and organizations 112as a profession 1buyouts 98, 171

Index

310

corporate 120development 232, 239

in BP 183–4, 239empowerment of 170in BP 183middle 237of modern firms 281–7of supply 214processes 17–18training 232

managerial vision see vision, managerialmanagers 1

and current results 148and interventions 99–100and organizational design 12communication among 237general 12motivation 161pay 151risk-taking and 120

manipulation of performance measures156, 159

Mannesmann 246, 251manufacturing:

lean 47, 49–51, 64modern 49 tpiece rates 148quality in 148

March, J. 254market exchanges 77market-facing units 168“markets for lemons” problem 89markets 22, 76–7, 84

and firms 88–103dealings 84, 199failure 78, 82–3, 105foreign 181for information 80organization 103transactions in 88

Martinez, J. 191mass production 47–50, 48 tMasten, S. E. 116 nmatrix forms 239–40

in ABB Asea Brown Boveri 239

system in BP 183Matsusaka, J. 214Matsushita 219McMillan, J. 77, 206, 242 nMcQuade, K. 209measures; see also performance, measures

of behavior 140of value 22results 140

Meckling, W. 103Meehan, J. W. 116 n, 117 n“merger of equals” 251–2mergers 226

and acquisitions 214, 247, 250–1merit pay 146; see also pay

for teachers in USA 146–7measurement of 147

Meyer, M. 225middle management 237middlemen 4, 8, 10milestones 270Milgrom, P. 31 n, 35, 41, 47, 73 n, 96,

99, 105, 106, 115 n, 116 n, 141,146, 169, 170, 177 n, 178 n, 201,225, 226, 233, 242 n

Mirrlees, J. 81mix and match 33, 241, 253modern manufacturing see

manufacturing, modernMohammed, A. 258monitoring 154

intermittent 124behavior 128, 157

of customers’ 192of employees’ 128, 172, 176

of Board 160of decision-making 171of employees 124

monopoly 79, 81, 88Monteverde, K. 116 nMontgomery, C. A. 214, 225Moore, J. 95, 115 nmoral hazard 83–4, 123, 220, 228motivation 75, 88, 118–79, 177 n, 234,

236, 270–1: see also PARC

Index

311

motivation (cont.):and organizational architecture

168in modern firms 77–8, 107, 118–79,

262, 270in markets 77of unit managers 270problems of 75–6, 268–9

Motorola 29, 274–6multibusiness firms 223–4multidivisional form 1–2, 93multi-tasking 97, 105–7, 140–54, 172,

179 n, 262–3, 265, 269–73exploitation 269–70exploration 269–70for job design 268–80in agency relationships 140–54in firms 168–9, 262–3in organizational design 106, 153incentive schemes 147

Murphy, K. 139, 160, 161, 242 nMyerson, R. 116 n

Nagar, V. 242 nnature of the firm 74–115Newell Co. 250–1Newman, P. C. 30 n“nexus of contracts” 104Nickerson, J. 73 nNike 191Nissan 40Nokia Corporation 28–30, 164, 274–80,

285acquisitions 224architecture of 276culture in 30, 173–5efficiency in 274exploitation in 280exploration in 278, 280growth in 276incentives in 277innovations in 274–5, 277managerial vision in 164objectives in 29organizational design in 276

pay in 277risk-taking in 285sales 275–6strategy in 29teamwork 276–7“value-based leadership” 276–7Ventures Organization (NVO) 278–9

non-concave performance 56 f: see alsoconcavities

non-concavities 33–4, 51–67: see alsoconvexities

and coherent points 59 f, 61, 61 fnon-contractual rewards 161non-convexities 33–4, 51–67non-excludability 79non-rivalry in consumption 79Nordstrom 285–6norms:

in firms 56of behavior 18, 23of culture 180of trust 26of sharing information 113of hard work 155, 277

North West Company (NWC) 4–11, 13,30, 60

and value chain 16customer-oriented strategy 6customers and 4information in 5organization in 231organizational innovations in 4recruitment in 5reward systems in 5strategic recognition in 62strategy in 15supplier structure in 4

Novartis 215Novo Nordisk

“facilitators” 113–14NWC see North West Company

(NWC)

O’Connell, J. 166Ofek, E. 225, 226

Index

312

Ohno, T. 64opportunity costs 90optimality, local 57O’Reilly, C. 252organizational architecture see

architecture, organizationalorganizational choices see choices,

organizationalorganizational design 1–30, 32–73,

112, 118, 140–1, 151, 153, 176,180–1, 221–2, 234, 243, 245,251–3, 255, 273, 280–1, 283–4,286–7

and influence activities 101and managers 11differentiation 221–2economics and 12for performance 191framework for 281in BP 185–6, 188, 190, 240, 288in firms 1–2, 101, 112, 118–19, 181,

281in Lincoln Electric Company 42, 64,

255–6in Nokia Corporation 276North West Company (NWC) 11in Novo Nordisk 113multi-tasking in 106–7, 273multidivisional 1–2patterns of 88problems of 13, 36, 153, 283–4strategy and 1–30, 19 f, 23–7, 32, 57,

61–4, 109, 281–3units 153, 167variation in 71

organizational exploitation 255–7, 262,264, 267–9, 271, 273

organizational exploration 255, 257–8,262–4, 267–9, 271–3

in Nokia Corporation 278, 280incentives for 266

organizations 10–11, 16–17, 55, 88,108–9, 114, 118, 224

and incentives 266and management 112

and strategy 1–30, 19 f, 32, 61–4,281–3

creativity 286behavior in 106–7, 224centralized decision-making 55changes in 2, 11, 23–7, 66–8, 231–2,

237, 241, 282compensations in 266complementarities and internal

organization 240decentralization in 93design of 180economics of 12excellence in 167growth in 243–80in BP 182internal 240

and performance 230–40managed 123motivation in 118–79of firms 180–241people in 17performance in 237strategy in 28, 109, 283structure of 1units in 231–32

leaders of 232visions in 175

organizing for growth 243–80organizing for performance see

performance, organizing foroutcomes-based rewards 126, 172outsourcing 2, 114, 170, 191–5,

199–200, 242 n, 250costs 195in BP 188in Toyota 203

ownership 105, 155and power 105of assets 104of firms 171of Lincoln Electric Company 44stakes 169structure of 171

Oyer, P. 156

Index

313

PARC framework 17, 254and motivation 164–76

partners 10, 201partnerships 2, 171, 198, 210–11, 213pay 136, 161, 175; see also merit pay;

performance-based incentive paydirect performance 167group performance 154in BP 238in Nokia Corporation 277managers 151performance 237stock-based 155teachers in USA 146–47withholding of 159

Pearce, D. 200Pearson, A. 111, 181peer:

assists see BP, peer assistsgroups see BP, peer groups

performance 3, 19–20, 37, 56–7, 68,129, 139, 142–4, 152, 180, 182,229–30, 233, 240, 255, 271: seealso organizations, internal andperformance

agent 149and changes 66

environmental 68and choice 52, 53 f, 54–5 f, 57, 59 fcollective 154contracts, design of 136

in BP 189–90corporate 155current 169, 243, 246, 264, 269economic 3evaluations 124, 139, 159–60

comparative 152in BP Exploration 112

in activities 144–5in firms 66, 68, 71, 139in jobs 114, 142–3in Lincoln Electric Company

144in organizations 237in units 233–4

incentives 99–100, 108, 152, 165,170, 232

investments in 97–8information on 158improvements of 182, 232linkages between 240measures 83, 94, 97–8, 106, 112, 130,

133–40, 142, 149–51, 154, 165,171–2, 175–6, 234–5, 240–1,262–3

in 3M 259in smaller units 233–4manipulation of 156of divisions 150–1of teachers in USA 147random 128, 234rewards and 172of colleagues, subversion of 154

of suppliers for Toyota 203organizational design for 191pay 85, 134–6, 140, 148, 237

direct 167group performance 154individual 154performance-based 176

rewards, withholding of 159targets in BP 186, 188test, of teachers in USA 147variation in 85

performance-based incentive pay 35, 176“petropreneurs” 185Pfeffer, J. 46Philips 219, 257piece rates 42–5, 139, 148

in Lincoln Electric Company 42–4,64, 256

in manufacturing 148in Safelite Auto Glass 165

Podolny, J. 13, 25, 113, 280 nPorter, M. 12postmerger integration 251

organization after 251power and ownership 105Prendergast, C. 140, 177 nPrennushi, G. 46

Index

314

processes, control of 255Proctor & Gamble 257

innovation in 266–7procurement 199production 167–8

in BP 185in Lincoln Electric Company 43lines in Japan 63

productivity 38, 45–6, 63in Lincoln Electric Company 45–6

products 47, 261development of 167–8

in Sony 220–1profitability 245property rights 95, 105public goods 79–80, 87

excludable 79–80public traded corporations 171purpose of the firm 74–115

quality:circles 63control 203in Lincoln Electric Company 144in manufacturing 148of goods 82

quasirents 92, 116 n

R&D groups 259in Proctor and Gamble 266

Rajan, R. 104, 225recruitment:

in North West Company (NWC) 5of employees 164–5

Reddy, P. 250relational contracts see contracts,

relationalrelationships between firms 209, 213

long-term 209, 213–14renegotiation:

of contracts 86–7, 92, 115 nof employment 85–7

rents 116 nreputations 121–2, 125, 160–4

and contracts 120–1

as motivation 161formation of 162in agency context 242 nin reward systems 163in Toyota 206in Vietnam 163

Research and Development groups 259results, measurement of 140rewards 124, 128, 130–1, 139, 142–4,

154–5, 158–9, 165for subjective evaluations 270–1in 3M 259non-contractual 161outcomes-based 126performance-based 144, 176

measures 176schemes 131, 140, 150, 154, 158–9

and behavior 159for sales staff 138formulaic 158

systems 138, 154, 165in North West Company (NWC) 5politicization of 160reputations in 163

risk 134–5, 151, 216, 234, 263, 266aversion 131–4, 165, 262, 265–6manage 255reduction 215, 242 nsharing 129–30taking 171, 268

and managers 120risk-neutrality 132, 178 nRoberts, J. 25, 28, 31 n, 35, 41, 47,

73 n, 99, 113, 115 n, 116 n, 141,146, 169, 170, 177 n, 178 n, 201,218, 225, 226, 233, 242 n, 280 n

Roos, D. 197Rosen, S. 139Rotemberg, J. 164Roy, D. 154Rubbermaid 250Ruigrok, W. 240

Sabel, C. 202sabotage of projects 224

Index

315

Safelite Auto Glass 165piece rates 165

salaries see paysales:

force 138, 149, 193–4in Nokia Corporation 276incentives 105–6, 149, 193–4

Saloner, G. 13, 164, 280 nSatterthwaite, M. 116 nSchaefer, S. 67, 177 n, 229Scharfstein, D. 225Schmittlein, D. C. 106, 149scope:

horizontal 214–30, 240vertical 214–30, 240

Sculley, J. 273Sears Roebuck 157–8selection 71–2selective intervention 94–9Servaes, H. 104, 225shareholders 120, 215, 243, 247

value 14, 181, 246, 249Shaw, K. 46Shepard, A. 13, 280 nShin, H.-H. 225short-term contracts see contracts,

short-termSimon, D. 184Simon, H. 77, 103“skunk works” model 271Smith, A. 74, 76Smith, J. 62: see also General Motors

(GM)Snyder, E. A. 116 n, 117 nSolectron 192Solow, R. 38Sony 218–23, 257

allocating resources 223acquisitions and 220and Columbia Pictures 218, 251and leadership 220–1and Philips 219control in 223differentiation in 222

influence activities in 223products in 220–1

specialized investments 90Spence, A. M. 82spillovers 73 n, 238spin-offs 114Spraakman, G. P. 31 nStacchetti, E. 200Stanford University Graduate School of

Business 168Stein, J. 225Stevenson, H. 191Stiglitz, J. 81stock:

market prices 137markets 214, 227, 247

in USA 225options 266prices and compensation 249, 270–1

stock-based pay 155strategic recognition 62strategy 12–15, 24, 57, 230, 255

and choice 32and North West Company (NWC) 15and organizations 1–30, 19 f, 23–7,

32, 57, 61–4, 109, 281–3creativity 286

clarity of 231, 234, 237, 240corporate 16, 237customer-oriented 6for growth 230in Nokia Corporation 29top-down 27

Stulz, R. M. 214, 225subjective evaluations 158–9

in 3M 259rewards for 270–1

suppliers:and General Motors (GM)

207–8in Toyota 202, 205

performance 203two-supplier policy 202–3

long-term 196 f, 200, 214

Index

316

relations with 2, 4, 200Toyota 204–5

structure in North West Company 4synergies 249–50

team:work and cooperation 154work in Nokia Corporation 276

team-organized work 42, 124Teece, D. 116 nTenaris 168

supply management system 168Tenneco 222–3, 252

compensation systems in 222corporate processes in 222

Thermo-Electron Corporation 95tight coupling 34, 67–73: see also loose

couplingTirole, J. 99, 170, 233Titman, S. 280 ntop-down strategy 27“tournaments” 139Toyoda, E. 64Toyota 39–40, 196, 201–6

and General Motors (GM) 201and suppliers 202–6

reputation 206two-supplier policy 202–3

assembly operation in USA 201cooperation in 201inventories in 65just-in-time in 65product line 39Production System (TPS) 64, 201reputations in 206suppliers’ performance for 203suppliers’ relations with 205

tracking stockstransaction costs 88, 90, 93, 105: see also

coststransfer:

and retention 71prices 102

two-supplier policy in Toyota 202–3

units:business, small 166

empowerment of 235–6, 239discrete business 189, 231

leaders of 232market-facing 168organizational 167performance of 151–3

UPF–Thompson 208

value 20–2, 96, 216, 228–30, 261and diversification 226corporate 285creation 13, 20, 104, 197–8, 245–7,

249–51, 270, 286in the firm 104, 228long-term 98maximization 20–2, 127measurement of 22of acquisitions 248shareholder 14statements of 285

value-chain 16organizer 191

van den Steen, E. 155, 164Vance, R. 153variation 71–2

in organizational design 71in performance 85

venture capitalists 172, 261vertical architect 191vertical disintegration 191vertical integration 93, 96, 117 n, 193,

195Viacom 246Vietnam, reputations in 163Villalonga, B. 226, 227Virgin 217vision 63

managerial 164of strategy and organization

284in organizations 175

Vodafone 246

Index

317

wages see payWaldman, M. 177 nWalt Disney Co. 218Wernerfeldt, B. 214, 225Westinghouse Electric 41, 246, 251Westinghouse/CBS 251Whang, S. 191Whinston, M. 117 nWhittington, R. 240, 242 nWilliamson, O. 86, 90, 94, 115 n, 216,

222Wilson, J. 153Wilson, R. 90“winners’ curse” 248

Womack, J. P. 197Woodruff, C. 163work-in-process 45, 68

in Lincoln Electric Company68, 256

Worldcom 156–8Wozniak, Steve 164Wulff, J. 242 n

Xerox 209–12, 254

Zemsky, P. 218Zenger, T. 73 nZingales, L. 104, 225

Index

318

the modern firm - department home · holmström, paul milgrom, and joel podolny. they will...

Documents